blob: 82e1063944bfc686da9a638531a00eda31a4b9ef [file] [log] [blame]
Willy Tarreaubaaee002006-06-26 02:48:02 +02001/*
2 * General purpose functions.
3 *
Willy Tarreau348238b2010-01-18 15:05:57 +01004 * Copyright 2000-2010 Willy Tarreau <w@1wt.eu>
Willy Tarreaubaaee002006-06-26 02:48:02 +02005 *
6 * This program is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU General Public License
8 * as published by the Free Software Foundation; either version
9 * 2 of the License, or (at your option) any later version.
10 *
11 */
12
Baruch Siache1651b22020-07-24 07:52:20 +030013#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +010014#define _GNU_SOURCE
15#include <dlfcn.h>
16#include <link.h>
17#endif
18
Willy Tarreau2e74c3f2007-12-02 18:45:09 +010019#include <ctype.h>
Willy Tarreau16e01562016-08-09 16:46:18 +020020#include <errno.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020021#include <netdb.h>
Willy Tarreau9a7bea52012-04-27 11:16:50 +020022#include <stdarg.h>
Willy Tarreaudd2f85e2012-09-02 22:34:23 +020023#include <stdio.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020024#include <stdlib.h>
25#include <string.h>
Thierry Fournier93127942016-01-20 18:49:45 +010026#include <time.h>
Willy Tarreau16e01562016-08-09 16:46:18 +020027#include <unistd.h>
Willy Tarreau127f9662007-12-06 00:53:51 +010028#include <sys/socket.h>
Willy Tarreau37101052019-05-20 16:48:20 +020029#include <sys/stat.h>
30#include <sys/types.h>
Willy Tarreau127f9662007-12-06 00:53:51 +010031#include <sys/un.h>
Willy Tarreaubaaee002006-06-26 02:48:02 +020032#include <netinet/in.h>
33#include <arpa/inet.h>
34
Willy Tarreau30053062020-08-20 16:39:14 +020035#if (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 16))
36#include <sys/auxv.h>
37#endif
38
Willy Tarreau48fbcae2020-06-03 18:09:46 +020039#include <import/eb32sctree.h>
Willy Tarreaub2551052020-06-09 09:07:15 +020040#include <import/eb32tree.h>
Willy Tarreau48fbcae2020-06-03 18:09:46 +020041
Willy Tarreau4c7e4b72020-05-27 12:58:42 +020042#include <haproxy/api.h>
Willy Tarreauc13ed532020-06-02 10:22:45 +020043#include <haproxy/chunk.h>
Willy Tarreau7c18b542020-06-11 09:23:02 +020044#include <haproxy/dgram.h>
Willy Tarreauf268ee82020-06-04 17:05:57 +020045#include <haproxy/global.h>
Willy Tarreau86416052020-06-04 09:20:54 +020046#include <haproxy/hlua.h>
Willy Tarreau213e9902020-06-04 14:58:24 +020047#include <haproxy/listener.h>
Willy Tarreau7a00efb2020-06-02 17:02:59 +020048#include <haproxy/namespace.h>
Christopher Faulet9553de72021-02-26 09:12:50 +010049#include <haproxy/net_helper.h>
Willy Tarreau5fc93282020-09-16 18:25:03 +020050#include <haproxy/protocol.h>
Emeric Brunc9437992021-02-12 19:42:55 +010051#include <haproxy/resolvers.h>
Willy Tarreau586f71b2020-12-11 15:54:36 +010052#include <haproxy/sock.h>
Willy Tarreau209108d2020-06-04 20:30:20 +020053#include <haproxy/ssl_sock.h>
Willy Tarreau5e539c92020-06-04 20:45:39 +020054#include <haproxy/stream_interface.h>
Willy Tarreaucea0e1b2020-06-04 17:25:40 +020055#include <haproxy/task.h>
Willy Tarreau48fbcae2020-06-03 18:09:46 +020056#include <haproxy/tools.h>
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +010057
Thierry Fournier93127942016-01-20 18:49:45 +010058/* This macro returns false if the test __x is false. Many
59 * of the following parsing function must be abort the processing
60 * if it returns 0, so this macro is useful for writing light code.
61 */
62#define RET0_UNLESS(__x) do { if (!(__x)) return 0; } while (0)
63
Willy Tarreau56adcf22012-12-23 18:00:29 +010064/* enough to store NB_ITOA_STR integers of :
Willy Tarreau72d759c2007-10-25 12:14:10 +020065 * 2^64-1 = 18446744073709551615 or
66 * -2^63 = -9223372036854775808
Willy Tarreaue7239b52009-03-29 13:41:58 +020067 *
68 * The HTML version needs room for adding the 25 characters
69 * '<span class="rls"></span>' around digits at positions 3N+1 in order
70 * to add spacing at up to 6 positions : 18 446 744 073 709 551 615
Willy Tarreau72d759c2007-10-25 12:14:10 +020071 */
Christopher Faulet99bca652017-11-14 16:47:26 +010072THREAD_LOCAL char itoa_str[NB_ITOA_STR][171];
73THREAD_LOCAL int itoa_idx = 0; /* index of next itoa_str to use */
Willy Tarreaubaaee002006-06-26 02:48:02 +020074
Willy Tarreau588297f2014-06-16 15:16:40 +020075/* sometimes we'll need to quote strings (eg: in stats), and we don't expect
76 * to quote strings larger than a max configuration line.
77 */
Christopher Faulet99bca652017-11-14 16:47:26 +010078THREAD_LOCAL char quoted_str[NB_QSTR][QSTR_SIZE + 1];
79THREAD_LOCAL int quoted_idx = 0;
Willy Tarreau588297f2014-06-16 15:16:40 +020080
Willy Tarreaubaaee002006-06-26 02:48:02 +020081/*
William Lallemande7340ec2012-01-24 11:15:39 +010082 * unsigned long long ASCII representation
83 *
84 * return the last char '\0' or NULL if no enough
85 * space in dst
86 */
87char *ulltoa(unsigned long long n, char *dst, size_t size)
88{
89 int i = 0;
90 char *res;
91
92 switch(n) {
93 case 1ULL ... 9ULL:
94 i = 0;
95 break;
96
97 case 10ULL ... 99ULL:
98 i = 1;
99 break;
100
101 case 100ULL ... 999ULL:
102 i = 2;
103 break;
104
105 case 1000ULL ... 9999ULL:
106 i = 3;
107 break;
108
109 case 10000ULL ... 99999ULL:
110 i = 4;
111 break;
112
113 case 100000ULL ... 999999ULL:
114 i = 5;
115 break;
116
117 case 1000000ULL ... 9999999ULL:
118 i = 6;
119 break;
120
121 case 10000000ULL ... 99999999ULL:
122 i = 7;
123 break;
124
125 case 100000000ULL ... 999999999ULL:
126 i = 8;
127 break;
128
129 case 1000000000ULL ... 9999999999ULL:
130 i = 9;
131 break;
132
133 case 10000000000ULL ... 99999999999ULL:
134 i = 10;
135 break;
136
137 case 100000000000ULL ... 999999999999ULL:
138 i = 11;
139 break;
140
141 case 1000000000000ULL ... 9999999999999ULL:
142 i = 12;
143 break;
144
145 case 10000000000000ULL ... 99999999999999ULL:
146 i = 13;
147 break;
148
149 case 100000000000000ULL ... 999999999999999ULL:
150 i = 14;
151 break;
152
153 case 1000000000000000ULL ... 9999999999999999ULL:
154 i = 15;
155 break;
156
157 case 10000000000000000ULL ... 99999999999999999ULL:
158 i = 16;
159 break;
160
161 case 100000000000000000ULL ... 999999999999999999ULL:
162 i = 17;
163 break;
164
165 case 1000000000000000000ULL ... 9999999999999999999ULL:
166 i = 18;
167 break;
168
169 case 10000000000000000000ULL ... ULLONG_MAX:
170 i = 19;
171 break;
172 }
173 if (i + 2 > size) // (i + 1) + '\0'
174 return NULL; // too long
175 res = dst + i + 1;
176 *res = '\0';
177 for (; i >= 0; i--) {
178 dst[i] = n % 10ULL + '0';
179 n /= 10ULL;
180 }
181 return res;
182}
183
184/*
185 * unsigned long ASCII representation
186 *
187 * return the last char '\0' or NULL if no enough
188 * space in dst
189 */
190char *ultoa_o(unsigned long n, char *dst, size_t size)
191{
192 int i = 0;
193 char *res;
194
195 switch (n) {
196 case 0U ... 9UL:
197 i = 0;
198 break;
199
200 case 10U ... 99UL:
201 i = 1;
202 break;
203
204 case 100U ... 999UL:
205 i = 2;
206 break;
207
208 case 1000U ... 9999UL:
209 i = 3;
210 break;
211
212 case 10000U ... 99999UL:
213 i = 4;
214 break;
215
216 case 100000U ... 999999UL:
217 i = 5;
218 break;
219
220 case 1000000U ... 9999999UL:
221 i = 6;
222 break;
223
224 case 10000000U ... 99999999UL:
225 i = 7;
226 break;
227
228 case 100000000U ... 999999999UL:
229 i = 8;
230 break;
231#if __WORDSIZE == 32
232
233 case 1000000000ULL ... ULONG_MAX:
234 i = 9;
235 break;
236
237#elif __WORDSIZE == 64
238
239 case 1000000000ULL ... 9999999999UL:
240 i = 9;
241 break;
242
243 case 10000000000ULL ... 99999999999UL:
244 i = 10;
245 break;
246
247 case 100000000000ULL ... 999999999999UL:
248 i = 11;
249 break;
250
251 case 1000000000000ULL ... 9999999999999UL:
252 i = 12;
253 break;
254
255 case 10000000000000ULL ... 99999999999999UL:
256 i = 13;
257 break;
258
259 case 100000000000000ULL ... 999999999999999UL:
260 i = 14;
261 break;
262
263 case 1000000000000000ULL ... 9999999999999999UL:
264 i = 15;
265 break;
266
267 case 10000000000000000ULL ... 99999999999999999UL:
268 i = 16;
269 break;
270
271 case 100000000000000000ULL ... 999999999999999999UL:
272 i = 17;
273 break;
274
275 case 1000000000000000000ULL ... 9999999999999999999UL:
276 i = 18;
277 break;
278
279 case 10000000000000000000ULL ... ULONG_MAX:
280 i = 19;
281 break;
282
283#endif
284 }
285 if (i + 2 > size) // (i + 1) + '\0'
286 return NULL; // too long
287 res = dst + i + 1;
288 *res = '\0';
289 for (; i >= 0; i--) {
290 dst[i] = n % 10U + '0';
291 n /= 10U;
292 }
293 return res;
294}
295
296/*
297 * signed long ASCII representation
298 *
299 * return the last char '\0' or NULL if no enough
300 * space in dst
301 */
302char *ltoa_o(long int n, char *dst, size_t size)
303{
304 char *pos = dst;
305
306 if (n < 0) {
307 if (size < 3)
308 return NULL; // min size is '-' + digit + '\0' but another test in ultoa
309 *pos = '-';
310 pos++;
311 dst = ultoa_o(-n, pos, size - 1);
312 } else {
313 dst = ultoa_o(n, dst, size);
314 }
315 return dst;
316}
317
318/*
319 * signed long long ASCII representation
320 *
321 * return the last char '\0' or NULL if no enough
322 * space in dst
323 */
324char *lltoa(long long n, char *dst, size_t size)
325{
326 char *pos = dst;
327
328 if (n < 0) {
329 if (size < 3)
330 return NULL; // min size is '-' + digit + '\0' but another test in ulltoa
331 *pos = '-';
332 pos++;
333 dst = ulltoa(-n, pos, size - 1);
334 } else {
335 dst = ulltoa(n, dst, size);
336 }
337 return dst;
338}
339
340/*
341 * write a ascii representation of a unsigned into dst,
342 * return a pointer to the last character
343 * Pad the ascii representation with '0', using size.
344 */
345char *utoa_pad(unsigned int n, char *dst, size_t size)
346{
347 int i = 0;
348 char *ret;
349
350 switch(n) {
351 case 0U ... 9U:
352 i = 0;
353 break;
354
355 case 10U ... 99U:
356 i = 1;
357 break;
358
359 case 100U ... 999U:
360 i = 2;
361 break;
362
363 case 1000U ... 9999U:
364 i = 3;
365 break;
366
367 case 10000U ... 99999U:
368 i = 4;
369 break;
370
371 case 100000U ... 999999U:
372 i = 5;
373 break;
374
375 case 1000000U ... 9999999U:
376 i = 6;
377 break;
378
379 case 10000000U ... 99999999U:
380 i = 7;
381 break;
382
383 case 100000000U ... 999999999U:
384 i = 8;
385 break;
386
387 case 1000000000U ... 4294967295U:
388 i = 9;
389 break;
390 }
391 if (i + 2 > size) // (i + 1) + '\0'
392 return NULL; // too long
393 if (i < size)
394 i = size - 2; // padding - '\0'
395
396 ret = dst + i + 1;
397 *ret = '\0';
398 for (; i >= 0; i--) {
399 dst[i] = n % 10U + '0';
400 n /= 10U;
401 }
402 return ret;
403}
404
405/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200406 * copies at most <size-1> chars from <src> to <dst>. Last char is always
407 * set to 0, unless <size> is 0. The number of chars copied is returned
408 * (excluding the terminating zero).
409 * This code has been optimized for size and speed : on x86, it's 45 bytes
410 * long, uses only registers, and consumes only 4 cycles per char.
411 */
412int strlcpy2(char *dst, const char *src, int size)
413{
414 char *orig = dst;
415 if (size) {
416 while (--size && (*dst = *src)) {
417 src++; dst++;
418 }
419 *dst = 0;
420 }
421 return dst - orig;
422}
423
424/*
Willy Tarreau72d759c2007-10-25 12:14:10 +0200425 * This function simply returns a locally allocated string containing
Willy Tarreaubaaee002006-06-26 02:48:02 +0200426 * the ascii representation for number 'n' in decimal.
427 */
Emeric Brun3a7fce52010-01-04 14:54:38 +0100428char *ultoa_r(unsigned long n, char *buffer, int size)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200429{
430 char *pos;
431
Willy Tarreau72d759c2007-10-25 12:14:10 +0200432 pos = buffer + size - 1;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200433 *pos-- = '\0';
434
435 do {
436 *pos-- = '0' + n % 10;
437 n /= 10;
Willy Tarreau72d759c2007-10-25 12:14:10 +0200438 } while (n && pos >= buffer);
Willy Tarreaubaaee002006-06-26 02:48:02 +0200439 return pos + 1;
440}
441
Willy Tarreau91092e52007-10-25 16:58:42 +0200442/*
Willy Tarreaue7239b52009-03-29 13:41:58 +0200443 * This function simply returns a locally allocated string containing
Thierry FOURNIER763a5d82015-07-06 23:09:52 +0200444 * the ascii representation for number 'n' in decimal.
445 */
446char *lltoa_r(long long int in, char *buffer, int size)
447{
448 char *pos;
449 int neg = 0;
450 unsigned long long int n;
451
452 pos = buffer + size - 1;
453 *pos-- = '\0';
454
455 if (in < 0) {
456 neg = 1;
457 n = -in;
458 }
459 else
460 n = in;
461
462 do {
463 *pos-- = '0' + n % 10;
464 n /= 10;
465 } while (n && pos >= buffer);
466 if (neg && pos > buffer)
467 *pos-- = '-';
468 return pos + 1;
469}
470
471/*
472 * This function simply returns a locally allocated string containing
Thierry FOURNIER1480bd82015-06-06 19:14:59 +0200473 * the ascii representation for signed number 'n' in decimal.
474 */
475char *sltoa_r(long n, char *buffer, int size)
476{
477 char *pos;
478
479 if (n >= 0)
480 return ultoa_r(n, buffer, size);
481
482 pos = ultoa_r(-n, buffer + 1, size - 1) - 1;
483 *pos = '-';
484 return pos;
485}
486
487/*
488 * This function simply returns a locally allocated string containing
Willy Tarreaue7239b52009-03-29 13:41:58 +0200489 * the ascii representation for number 'n' in decimal, formatted for
490 * HTML output with tags to create visual grouping by 3 digits. The
491 * output needs to support at least 171 characters.
492 */
493const char *ulltoh_r(unsigned long long n, char *buffer, int size)
494{
495 char *start;
496 int digit = 0;
497
498 start = buffer + size;
499 *--start = '\0';
500
501 do {
502 if (digit == 3 && start >= buffer + 7)
503 memcpy(start -= 7, "</span>", 7);
504
505 if (start >= buffer + 1) {
506 *--start = '0' + n % 10;
507 n /= 10;
508 }
509
510 if (digit == 3 && start >= buffer + 18)
511 memcpy(start -= 18, "<span class=\"rls\">", 18);
512
513 if (digit++ == 3)
514 digit = 1;
515 } while (n && start > buffer);
516 return start;
517}
518
519/*
Willy Tarreau91092e52007-10-25 16:58:42 +0200520 * This function simply returns a locally allocated string containing the ascii
521 * representation for number 'n' in decimal, unless n is 0 in which case it
522 * returns the alternate string (or an empty string if the alternate string is
523 * NULL). It use is intended for limits reported in reports, where it's
524 * desirable not to display anything if there is no limit. Warning! it shares
525 * the same vector as ultoa_r().
526 */
527const char *limit_r(unsigned long n, char *buffer, int size, const char *alt)
528{
529 return (n) ? ultoa_r(n, buffer, size) : (alt ? alt : "");
530}
531
Willy Tarreau588297f2014-06-16 15:16:40 +0200532/* returns a locally allocated string containing the quoted encoding of the
533 * input string. The output may be truncated to QSTR_SIZE chars, but it is
534 * guaranteed that the string will always be properly terminated. Quotes are
535 * encoded by doubling them as is commonly done in CSV files. QSTR_SIZE must
536 * always be at least 4 chars.
537 */
538const char *qstr(const char *str)
539{
540 char *ret = quoted_str[quoted_idx];
541 char *p, *end;
542
543 if (++quoted_idx >= NB_QSTR)
544 quoted_idx = 0;
545
546 p = ret;
547 end = ret + QSTR_SIZE;
548
549 *p++ = '"';
550
551 /* always keep 3 chars to support passing "" and the ending " */
552 while (*str && p < end - 3) {
553 if (*str == '"') {
554 *p++ = '"';
555 *p++ = '"';
556 }
557 else
558 *p++ = *str;
559 str++;
560 }
561 *p++ = '"';
562 return ret;
563}
564
Robert Tsai81ae1952007-12-05 10:47:29 +0100565/*
Willy Tarreaubaaee002006-06-26 02:48:02 +0200566 * Returns non-zero if character <s> is a hex digit (0-9, a-f, A-F), else zero.
567 *
568 * It looks like this one would be a good candidate for inlining, but this is
569 * not interesting because it around 35 bytes long and often called multiple
570 * times within the same function.
571 */
572int ishex(char s)
573{
574 s -= '0';
575 if ((unsigned char)s <= 9)
576 return 1;
577 s -= 'A' - '0';
578 if ((unsigned char)s <= 5)
579 return 1;
580 s -= 'a' - 'A';
581 if ((unsigned char)s <= 5)
582 return 1;
583 return 0;
584}
585
Willy Tarreau3ca1a882015-01-15 18:43:49 +0100586/* rounds <i> down to the closest value having max 2 digits */
587unsigned int round_2dig(unsigned int i)
588{
589 unsigned int mul = 1;
590
591 while (i >= 100) {
592 i /= 10;
593 mul *= 10;
594 }
595 return i * mul;
596}
597
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100598/*
599 * Checks <name> for invalid characters. Valid chars are [A-Za-z0-9_:.-]. If an
600 * invalid character is found, a pointer to it is returned. If everything is
601 * fine, NULL is returned.
602 */
603const char *invalid_char(const char *name)
604{
605 if (!*name)
606 return name;
607
608 while (*name) {
Willy Tarreau90807112020-02-25 08:16:33 +0100609 if (!isalnum((unsigned char)*name) && *name != '.' && *name != ':' &&
Willy Tarreau2e74c3f2007-12-02 18:45:09 +0100610 *name != '_' && *name != '-')
611 return name;
612 name++;
613 }
614 return NULL;
615}
Willy Tarreaubaaee002006-06-26 02:48:02 +0200616
617/*
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200618 * Checks <name> for invalid characters. Valid chars are [_.-] and those
619 * accepted by <f> function.
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200620 * If an invalid character is found, a pointer to it is returned.
621 * If everything is fine, NULL is returned.
622 */
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200623static inline const char *__invalid_char(const char *name, int (*f)(int)) {
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200624
625 if (!*name)
626 return name;
627
628 while (*name) {
Willy Tarreau90807112020-02-25 08:16:33 +0100629 if (!f((unsigned char)*name) && *name != '.' &&
Krzysztof Piotr Oledzkiefe3b6f2008-05-23 23:49:32 +0200630 *name != '_' && *name != '-')
631 return name;
632
633 name++;
634 }
635
636 return NULL;
637}
638
639/*
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200640 * Checks <name> for invalid characters. Valid chars are [A-Za-z0-9_.-].
641 * If an invalid character is found, a pointer to it is returned.
642 * If everything is fine, NULL is returned.
643 */
644const char *invalid_domainchar(const char *name) {
645 return __invalid_char(name, isalnum);
646}
647
648/*
649 * Checks <name> for invalid characters. Valid chars are [A-Za-z_.-].
650 * If an invalid character is found, a pointer to it is returned.
651 * If everything is fine, NULL is returned.
652 */
653const char *invalid_prefix_char(const char *name) {
Thierry Fournierf7b7c3e2018-03-26 11:54:39 +0200654 return __invalid_char(name, isalnum);
Frédéric Lécailleb82f7422017-04-13 18:24:23 +0200655}
656
657/*
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100658 * converts <str> to a struct sockaddr_storage* provided by the caller. The
Willy Tarreau24709282013-03-10 21:32:12 +0100659 * caller must have zeroed <sa> first, and may have set sa->ss_family to force
660 * parse a specific address format. If the ss_family is 0 or AF_UNSPEC, then
661 * the function tries to guess the address family from the syntax. If the
662 * family is forced and the format doesn't match, an error is returned. The
Willy Tarreaufab5a432011-03-04 15:31:53 +0100663 * string is assumed to contain only an address, no port. The address can be a
664 * dotted IPv4 address, an IPv6 address, a host name, or empty or "*" to
665 * indicate INADDR_ANY. NULL is returned if the host part cannot be resolved.
666 * The return address will only have the address family and the address set,
667 * all other fields remain zero. The string is not supposed to be modified.
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100668 * The IPv6 '::' address is IN6ADDR_ANY. If <resolve> is non-zero, the hostname
669 * is resolved, otherwise only IP addresses are resolved, and anything else
Willy Tarreauecde7df2016-11-02 22:37:03 +0100670 * returns NULL. If the address contains a port, this one is preserved.
Willy Tarreaubaaee002006-06-26 02:48:02 +0200671 */
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100672struct sockaddr_storage *str2ip2(const char *str, struct sockaddr_storage *sa, int resolve)
Willy Tarreaubaaee002006-06-26 02:48:02 +0200673{
Willy Tarreaufab5a432011-03-04 15:31:53 +0100674 struct hostent *he;
mildisff5d5102015-10-26 18:50:08 +0100675 /* max IPv6 length, including brackets and terminating NULL */
676 char tmpip[48];
Willy Tarreauecde7df2016-11-02 22:37:03 +0100677 int port = get_host_port(sa);
mildisff5d5102015-10-26 18:50:08 +0100678
679 /* check IPv6 with square brackets */
680 if (str[0] == '[') {
681 size_t iplength = strlen(str);
682
683 if (iplength < 4) {
684 /* minimal size is 4 when using brackets "[::]" */
685 goto fail;
686 }
687 else if (iplength >= sizeof(tmpip)) {
688 /* IPv6 literal can not be larger than tmpip */
689 goto fail;
690 }
691 else {
692 if (str[iplength - 1] != ']') {
693 /* if address started with bracket, it should end with bracket */
694 goto fail;
695 }
696 else {
697 memcpy(tmpip, str + 1, iplength - 2);
698 tmpip[iplength - 2] = '\0';
699 str = tmpip;
700 }
701 }
702 }
Willy Tarreaufab5a432011-03-04 15:31:53 +0100703
Willy Tarreaufab5a432011-03-04 15:31:53 +0100704 /* Any IPv6 address */
705 if (str[0] == ':' && str[1] == ':' && !str[2]) {
Willy Tarreau24709282013-03-10 21:32:12 +0100706 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
707 sa->ss_family = AF_INET6;
708 else if (sa->ss_family != AF_INET6)
709 goto fail;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100710 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100711 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100712 }
713
Willy Tarreau24709282013-03-10 21:32:12 +0100714 /* Any address for the family, defaults to IPv4 */
Willy Tarreaufab5a432011-03-04 15:31:53 +0100715 if (!str[0] || (str[0] == '*' && !str[1])) {
Willy Tarreau24709282013-03-10 21:32:12 +0100716 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
717 sa->ss_family = AF_INET;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100718 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100719 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100720 }
721
722 /* check for IPv6 first */
Willy Tarreau24709282013-03-10 21:32:12 +0100723 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET6) &&
724 inet_pton(AF_INET6, str, &((struct sockaddr_in6 *)sa)->sin6_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100725 sa->ss_family = AF_INET6;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100726 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100727 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100728 }
729
730 /* then check for IPv4 */
Willy Tarreau24709282013-03-10 21:32:12 +0100731 if ((!sa->ss_family || sa->ss_family == AF_UNSPEC || sa->ss_family == AF_INET) &&
732 inet_pton(AF_INET, str, &((struct sockaddr_in *)sa)->sin_addr)) {
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100733 sa->ss_family = AF_INET;
Willy Tarreauecde7df2016-11-02 22:37:03 +0100734 set_host_port(sa, port);
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100735 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100736 }
737
Thierry FOURNIER58639a02014-11-25 12:02:25 +0100738 if (!resolve)
739 return NULL;
740
Emeric Brund30e9a12020-12-23 18:49:16 +0100741 if (!resolv_hostname_validation(str, NULL))
Baptiste Assmanna68ca962015-04-14 01:15:08 +0200742 return NULL;
743
David du Colombierd5f43282011-03-17 10:40:16 +0100744#ifdef USE_GETADDRINFO
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200745 if (global.tune.options & GTUNE_USE_GAI) {
David du Colombierd5f43282011-03-17 10:40:16 +0100746 struct addrinfo hints, *result;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100747 int success = 0;
David du Colombierd5f43282011-03-17 10:40:16 +0100748
749 memset(&result, 0, sizeof(result));
750 memset(&hints, 0, sizeof(hints));
Willy Tarreau24709282013-03-10 21:32:12 +0100751 hints.ai_family = sa->ss_family ? sa->ss_family : AF_UNSPEC;
David du Colombierd5f43282011-03-17 10:40:16 +0100752 hints.ai_socktype = SOCK_DGRAM;
Dmitry Sivachenkoeab7f392015-10-02 01:01:58 +0200753 hints.ai_flags = 0;
David du Colombierd5f43282011-03-17 10:40:16 +0100754 hints.ai_protocol = 0;
755
756 if (getaddrinfo(str, NULL, &hints, &result) == 0) {
Willy Tarreau24709282013-03-10 21:32:12 +0100757 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
758 sa->ss_family = result->ai_family;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100759 else if (sa->ss_family != result->ai_family) {
760 freeaddrinfo(result);
Willy Tarreau24709282013-03-10 21:32:12 +0100761 goto fail;
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100762 }
Willy Tarreau24709282013-03-10 21:32:12 +0100763
David du Colombierd5f43282011-03-17 10:40:16 +0100764 switch (result->ai_family) {
765 case AF_INET:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100766 memcpy((struct sockaddr_in *)sa, result->ai_addr, result->ai_addrlen);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100767 set_host_port(sa, port);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100768 success = 1;
769 break;
David du Colombierd5f43282011-03-17 10:40:16 +0100770 case AF_INET6:
Willy Tarreauc120c8d2013-03-10 19:27:44 +0100771 memcpy((struct sockaddr_in6 *)sa, result->ai_addr, result->ai_addrlen);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100772 set_host_port(sa, port);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100773 success = 1;
774 break;
David du Colombierd5f43282011-03-17 10:40:16 +0100775 }
776 }
777
Sean Carey58ea0392013-02-15 23:39:18 +0100778 if (result)
779 freeaddrinfo(result);
Tim Duesterhus7d58b4d2018-01-21 22:11:17 +0100780
781 if (success)
782 return sa;
Willy Tarreaufab5a432011-03-04 15:31:53 +0100783 }
David du Colombierd5f43282011-03-17 10:40:16 +0100784#endif
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200785 /* try to resolve an IPv4/IPv6 hostname */
786 he = gethostbyname(str);
787 if (he) {
788 if (!sa->ss_family || sa->ss_family == AF_UNSPEC)
789 sa->ss_family = he->h_addrtype;
790 else if (sa->ss_family != he->h_addrtype)
791 goto fail;
792
793 switch (sa->ss_family) {
794 case AF_INET:
795 ((struct sockaddr_in *)sa)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100796 set_host_port(sa, port);
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200797 return sa;
798 case AF_INET6:
799 ((struct sockaddr_in6 *)sa)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
Willy Tarreauecde7df2016-11-02 22:37:03 +0100800 set_host_port(sa, port);
Nenad Merdanovic88afe032014-04-14 15:56:58 +0200801 return sa;
802 }
803 }
804
David du Colombierd5f43282011-03-17 10:40:16 +0100805 /* unsupported address family */
Willy Tarreau24709282013-03-10 21:32:12 +0100806 fail:
Willy Tarreaufab5a432011-03-04 15:31:53 +0100807 return NULL;
808}
809
810/*
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100811 * Converts <str> to a locally allocated struct sockaddr_storage *, and a port
812 * range or offset consisting in two integers that the caller will have to
813 * check to find the relevant input format. The following format are supported :
814 *
815 * String format | address | port | low | high
816 * addr | <addr> | 0 | 0 | 0
817 * addr: | <addr> | 0 | 0 | 0
818 * addr:port | <addr> | <port> | <port> | <port>
819 * addr:pl-ph | <addr> | <pl> | <pl> | <ph>
820 * addr:+port | <addr> | <port> | 0 | <port>
821 * addr:-port | <addr> |-<port> | <port> | 0
822 *
823 * The detection of a port range or increment by the caller is made by
824 * comparing <low> and <high>. If both are equal, then port 0 means no port
825 * was specified. The caller may pass NULL for <low> and <high> if it is not
826 * interested in retrieving port ranges.
827 *
828 * Note that <addr> above may also be :
829 * - empty ("") => family will be AF_INET and address will be INADDR_ANY
830 * - "*" => family will be AF_INET and address will be INADDR_ANY
831 * - "::" => family will be AF_INET6 and address will be IN6ADDR_ANY
832 * - a host name => family and address will depend on host name resolving.
833 *
Willy Tarreau24709282013-03-10 21:32:12 +0100834 * A prefix may be passed in before the address above to force the family :
835 * - "ipv4@" => force address to resolve as IPv4 and fail if not possible.
836 * - "ipv6@" => force address to resolve as IPv6 and fail if not possible.
837 * - "unix@" => force address to be a path to a UNIX socket even if the
838 * path does not start with a '/'
Willy Tarreauccfccef2014-05-10 01:49:15 +0200839 * - 'abns@' -> force address to belong to the abstract namespace (Linux
840 * only). These sockets are just like Unix sockets but without
841 * the need for an underlying file system. The address is a
842 * string. Technically it's like a Unix socket with a zero in
843 * the first byte of the address.
Willy Tarreau40aa0702013-03-10 23:51:38 +0100844 * - "fd@" => an integer must follow, and is a file descriptor number.
Willy Tarreau24709282013-03-10 21:32:12 +0100845 *
mildisff5d5102015-10-26 18:50:08 +0100846 * IPv6 addresses can be declared with or without square brackets. When using
847 * square brackets for IPv6 addresses, the port separator (colon) is optional.
848 * If not using square brackets, and in order to avoid any ambiguity with
849 * IPv6 addresses, the last colon ':' is mandatory even when no port is specified.
850 * NULL is returned if the address cannot be parsed. The <low> and <high> ports
851 * are always initialized if non-null, even for non-IP families.
Willy Tarreaud393a622013-03-04 18:22:00 +0100852 *
853 * If <pfx> is non-null, it is used as a string prefix before any path-based
854 * address (typically the path to a unix socket).
Willy Tarreau40aa0702013-03-10 23:51:38 +0100855 *
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200856 * if <fqdn> is non-null, it will be filled with :
857 * - a pointer to the FQDN of the server name to resolve if there's one, and
858 * that the caller will have to free(),
859 * - NULL if there was an explicit address that doesn't require resolution.
860 *
Willy Tarreaucd3a55912020-09-04 15:30:46 +0200861 * Hostnames are only resolved if <opts> has PA_O_RESOLVE. Otherwise <fqdn> is
862 * still honored so it is possible for the caller to know whether a resolution
863 * failed by clearing this flag and checking if <fqdn> was filled, indicating
864 * the need for a resolution.
Thierry FOURNIER7fe3be72015-09-26 20:03:36 +0200865 *
Willy Tarreau40aa0702013-03-10 23:51:38 +0100866 * When a file descriptor is passed, its value is put into the s_addr part of
Willy Tarreaua5b325f2020-09-04 16:44:20 +0200867 * the address when cast to sockaddr_in and the address family is
868 * AF_CUST_EXISTING_FD.
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200869 *
Willy Tarreau5fc93282020-09-16 18:25:03 +0200870 * The matching protocol will be set into <proto> if non-null.
871 *
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200872 * Any known file descriptor is also assigned to <fd> if non-null, otherwise it
873 * is forced to -1.
Willy Tarreaufab5a432011-03-04 15:31:53 +0100874 */
Willy Tarreau5fc93282020-09-16 18:25:03 +0200875struct sockaddr_storage *str2sa_range(const char *str, int *port, int *low, int *high, int *fd,
876 struct protocol **proto, char **err,
877 const char *pfx, char **fqdn, unsigned int opts)
Willy Tarreaufab5a432011-03-04 15:31:53 +0100878{
Christopher Faulet1bc04c72017-10-29 20:14:08 +0100879 static THREAD_LOCAL struct sockaddr_storage ss;
David du Colombier6f5ccb12011-03-10 22:26:24 +0100880 struct sockaddr_storage *ret = NULL;
Willy Tarreau5fc93282020-09-16 18:25:03 +0200881 struct protocol *new_proto = NULL;
Willy Tarreau24709282013-03-10 21:32:12 +0100882 char *back, *str2;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100883 char *port1, *port2;
884 int portl, porth, porta;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200885 int abstract = 0;
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200886 int new_fd = -1;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200887 int sock_type, ctrl_type;
Willy Tarreaud4448bc2013-02-20 15:55:15 +0100888
889 portl = porth = porta = 0;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +0200890 if (fqdn)
891 *fqdn = NULL;
Willy Tarreaubaaee002006-06-26 02:48:02 +0200892
Willy Tarreaudad36a32013-03-11 01:20:04 +0100893 str2 = back = env_expand(strdup(str));
Willy Tarreaudf350f12013-03-01 20:22:54 +0100894 if (str2 == NULL) {
895 memprintf(err, "out of memory in '%s'\n", __FUNCTION__);
Willy Tarreaud5191e72010-02-09 20:50:45 +0100896 goto out;
Willy Tarreaudf350f12013-03-01 20:22:54 +0100897 }
Willy Tarreaubaaee002006-06-26 02:48:02 +0200898
Willy Tarreau9f69f462015-09-08 16:01:25 +0200899 if (!*str2) {
900 memprintf(err, "'%s' resolves to an empty address (environment variable missing?)\n", str);
901 goto out;
902 }
903
Willy Tarreau24709282013-03-10 21:32:12 +0100904 memset(&ss, 0, sizeof(ss));
905
Willy Tarreaue835bd82020-09-16 11:35:47 +0200906 /* prepare the default socket types */
907 if ((opts & (PA_O_STREAM|PA_O_DGRAM)) == PA_O_DGRAM)
908 sock_type = ctrl_type = SOCK_DGRAM;
909 else
910 sock_type = ctrl_type = SOCK_STREAM;
911
912 if (strncmp(str2, "stream+", 7) == 0) {
913 str2 += 7;
914 sock_type = ctrl_type = SOCK_STREAM;
915 }
916 else if (strncmp(str2, "dgram+", 6) == 0) {
917 str2 += 6;
918 sock_type = ctrl_type = SOCK_DGRAM;
919 }
920
Willy Tarreau24709282013-03-10 21:32:12 +0100921 if (strncmp(str2, "unix@", 5) == 0) {
922 str2 += 5;
Willy Tarreauccfccef2014-05-10 01:49:15 +0200923 abstract = 0;
Willy Tarreau24709282013-03-10 21:32:12 +0100924 ss.ss_family = AF_UNIX;
925 }
Willy Tarreauccfccef2014-05-10 01:49:15 +0200926 else if (strncmp(str2, "abns@", 5) == 0) {
927 str2 += 5;
928 abstract = 1;
929 ss.ss_family = AF_UNIX;
930 }
Willy Tarreau24709282013-03-10 21:32:12 +0100931 else if (strncmp(str2, "ipv4@", 5) == 0) {
932 str2 += 5;
933 ss.ss_family = AF_INET;
934 }
935 else if (strncmp(str2, "ipv6@", 5) == 0) {
936 str2 += 5;
937 ss.ss_family = AF_INET6;
938 }
Emeric Brun3835c0d2020-07-07 09:46:09 +0200939 else if (strncmp(str2, "udp4@", 5) == 0) {
940 str2 += 5;
941 ss.ss_family = AF_INET;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200942 sock_type = ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +0200943 }
944 else if (strncmp(str2, "udp6@", 5) == 0) {
945 str2 += 5;
946 ss.ss_family = AF_INET6;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200947 sock_type = ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +0200948 }
949 else if (strncmp(str2, "udp@", 4) == 0) {
950 str2 += 4;
951 ss.ss_family = AF_UNSPEC;
Willy Tarreaue835bd82020-09-16 11:35:47 +0200952 sock_type = ctrl_type = SOCK_DGRAM;
Emeric Brun3835c0d2020-07-07 09:46:09 +0200953 }
Frédéric Lécaille10caf652020-11-23 11:36:57 +0100954 else if (strncmp(str2, "quic4@", 6) == 0) {
955 str2 += 6;
956 ss.ss_family = AF_INET;
957 sock_type = SOCK_DGRAM;
958 ctrl_type = SOCK_STREAM;
959 }
960 else if (strncmp(str2, "quic6@", 6) == 0) {
961 str2 += 6;
962 ss.ss_family = AF_INET6;
963 sock_type = SOCK_DGRAM;
964 ctrl_type = SOCK_STREAM;
965 }
Willy Tarreau5a7beed2020-09-04 16:54:05 +0200966 else if (strncmp(str2, "fd@", 3) == 0) {
967 str2 += 3;
968 ss.ss_family = AF_CUST_EXISTING_FD;
969 }
970 else if (strncmp(str2, "sockpair@", 9) == 0) {
971 str2 += 9;
972 ss.ss_family = AF_CUST_SOCKPAIR;
973 }
Willy Tarreau24709282013-03-10 21:32:12 +0100974 else if (*str2 == '/') {
975 ss.ss_family = AF_UNIX;
976 }
977 else
978 ss.ss_family = AF_UNSPEC;
979
Willy Tarreau5a7beed2020-09-04 16:54:05 +0200980 if (ss.ss_family == AF_CUST_SOCKPAIR) {
Willy Tarreaua215be22020-09-16 10:14:16 +0200981 struct sockaddr_storage ss2;
982 socklen_t addr_len;
William Lallemand2fe7dd02018-09-11 16:51:29 +0200983 char *endptr;
984
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200985 new_fd = strtol(str2, &endptr, 10);
986 if (!*str2 || new_fd < 0 || *endptr) {
William Lallemand2fe7dd02018-09-11 16:51:29 +0200987 memprintf(err, "file descriptor '%s' is not a valid integer in '%s'\n", str2, str);
988 goto out;
989 }
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200990
Willy Tarreaua215be22020-09-16 10:14:16 +0200991 /* just verify that it's a socket */
992 addr_len = sizeof(ss2);
993 if (getsockname(new_fd, (struct sockaddr *)&ss2, &addr_len) == -1) {
994 memprintf(err, "cannot use file descriptor '%d' : %s.\n", new_fd, strerror(errno));
995 goto out;
996 }
997
Willy Tarreaua93e5c72020-09-15 14:01:16 +0200998 ((struct sockaddr_in *)&ss)->sin_addr.s_addr = new_fd;
999 ((struct sockaddr_in *)&ss)->sin_port = 0;
William Lallemand2fe7dd02018-09-11 16:51:29 +02001000 }
Willy Tarreau5a7beed2020-09-04 16:54:05 +02001001 else if (ss.ss_family == AF_CUST_EXISTING_FD) {
Willy Tarreau40aa0702013-03-10 23:51:38 +01001002 char *endptr;
1003
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001004 new_fd = strtol(str2, &endptr, 10);
1005 if (!*str2 || new_fd < 0 || *endptr) {
Willy Tarreaudad36a32013-03-11 01:20:04 +01001006 memprintf(err, "file descriptor '%s' is not a valid integer in '%s'\n", str2, str);
Willy Tarreau40aa0702013-03-10 23:51:38 +01001007 goto out;
1008 }
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001009
Willy Tarreau6edc7222020-09-15 17:41:56 +02001010 if (opts & PA_O_SOCKET_FD) {
1011 socklen_t addr_len;
1012 int type;
1013
1014 addr_len = sizeof(ss);
1015 if (getsockname(new_fd, (struct sockaddr *)&ss, &addr_len) == -1) {
1016 memprintf(err, "cannot use file descriptor '%d' : %s.\n", new_fd, strerror(errno));
1017 goto out;
1018 }
1019
1020 addr_len = sizeof(type);
1021 if (getsockopt(new_fd, SOL_SOCKET, SO_TYPE, &type, &addr_len) != 0 ||
Willy Tarreaue835bd82020-09-16 11:35:47 +02001022 (type == SOCK_STREAM) != (sock_type == SOCK_STREAM)) {
Willy Tarreau6edc7222020-09-15 17:41:56 +02001023 memprintf(err, "socket on file descriptor '%d' is of the wrong type.\n", new_fd);
1024 goto out;
1025 }
1026
1027 porta = portl = porth = get_host_port(&ss);
1028 } else if (opts & PA_O_RAW_FD) {
1029 ((struct sockaddr_in *)&ss)->sin_addr.s_addr = new_fd;
1030 ((struct sockaddr_in *)&ss)->sin_port = 0;
1031 } else {
1032 memprintf(err, "a file descriptor is not acceptable here in '%s'\n", str);
1033 goto out;
1034 }
Willy Tarreau40aa0702013-03-10 23:51:38 +01001035 }
1036 else if (ss.ss_family == AF_UNIX) {
Willy Tarreau8daa9202019-06-16 18:16:33 +02001037 struct sockaddr_un *un = (struct sockaddr_un *)&ss;
Willy Tarreau15586382013-03-04 19:48:14 +01001038 int prefix_path_len;
1039 int max_path_len;
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001040 int adr_len;
Willy Tarreau15586382013-03-04 19:48:14 +01001041
1042 /* complete unix socket path name during startup or soft-restart is
1043 * <unix_bind_prefix><path>.<pid>.<bak|tmp>
1044 */
Willy Tarreauccfccef2014-05-10 01:49:15 +02001045 prefix_path_len = (pfx && !abstract) ? strlen(pfx) : 0;
Willy Tarreau8daa9202019-06-16 18:16:33 +02001046 max_path_len = (sizeof(un->sun_path) - 1) -
Willy Tarreau327ea5a2020-02-11 06:43:37 +01001047 (abstract ? 0 : prefix_path_len + 1 + 5 + 1 + 3);
Willy Tarreau15586382013-03-04 19:48:14 +01001048
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001049 adr_len = strlen(str2);
1050 if (adr_len > max_path_len) {
Willy Tarreau15586382013-03-04 19:48:14 +01001051 memprintf(err, "socket path '%s' too long (max %d)\n", str, max_path_len);
1052 goto out;
1053 }
1054
Willy Tarreauccfccef2014-05-10 01:49:15 +02001055 /* when abstract==1, we skip the first zero and copy all bytes except the trailing zero */
Willy Tarreau8daa9202019-06-16 18:16:33 +02001056 memset(un->sun_path, 0, sizeof(un->sun_path));
Willy Tarreau94ef3f32014-04-14 14:49:00 +02001057 if (prefix_path_len)
Willy Tarreau8daa9202019-06-16 18:16:33 +02001058 memcpy(un->sun_path, pfx, prefix_path_len);
1059 memcpy(un->sun_path + prefix_path_len + abstract, str2, adr_len + 1 - abstract);
Willy Tarreau15586382013-03-04 19:48:14 +01001060 }
Willy Tarreau24709282013-03-10 21:32:12 +01001061 else { /* IPv4 and IPv6 */
mildisff5d5102015-10-26 18:50:08 +01001062 char *end = str2 + strlen(str2);
1063 char *chr;
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001064
mildisff5d5102015-10-26 18:50:08 +01001065 /* search for : or ] whatever comes first */
1066 for (chr = end-1; chr > str2; chr--) {
1067 if (*chr == ']' || *chr == ':')
1068 break;
1069 }
1070
1071 if (*chr == ':') {
1072 /* Found a colon before a closing-bracket, must be a port separator.
1073 * This guarantee backward compatibility.
1074 */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001075 if (!(opts & PA_O_PORT_OK)) {
1076 memprintf(err, "port specification not permitted here in '%s'", str);
1077 goto out;
1078 }
mildisff5d5102015-10-26 18:50:08 +01001079 *chr++ = '\0';
1080 port1 = chr;
1081 }
1082 else {
1083 /* Either no colon and no closing-bracket
1084 * or directly ending with a closing-bracket.
1085 * However, no port.
1086 */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001087 if (opts & PA_O_PORT_MAND) {
1088 memprintf(err, "missing port specification in '%s'", str);
1089 goto out;
1090 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001091 port1 = "";
mildisff5d5102015-10-26 18:50:08 +01001092 }
Willy Tarreaubaaee002006-06-26 02:48:02 +02001093
Willy Tarreau90807112020-02-25 08:16:33 +01001094 if (isdigit((unsigned char)*port1)) { /* single port or range */
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001095 port2 = strchr(port1, '-');
Willy Tarreau7f96a842020-09-15 11:12:44 +02001096 if (port2) {
1097 if (!(opts & PA_O_PORT_RANGE)) {
1098 memprintf(err, "port range not permitted here in '%s'", str);
1099 goto out;
1100 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001101 *port2++ = '\0';
Willy Tarreau7f96a842020-09-15 11:12:44 +02001102 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001103 else
1104 port2 = port1;
1105 portl = atoi(port1);
1106 porth = atoi(port2);
Willy Tarreau7f96a842020-09-15 11:12:44 +02001107
1108 if (portl < !!(opts & PA_O_PORT_MAND) || portl > 65535) {
1109 memprintf(err, "invalid port '%s'", port1);
1110 goto out;
1111 }
1112
1113 if (porth < !!(opts & PA_O_PORT_MAND) || porth > 65535) {
1114 memprintf(err, "invalid port '%s'", port2);
1115 goto out;
1116 }
1117
1118 if (portl > porth) {
1119 memprintf(err, "invalid port range '%d-%d'", portl, porth);
1120 goto out;
1121 }
1122
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001123 porta = portl;
1124 }
1125 else if (*port1 == '-') { /* negative offset */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001126 if (!(opts & PA_O_PORT_OFS)) {
1127 memprintf(err, "port offset not permitted here in '%s'", str);
1128 goto out;
1129 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001130 portl = atoi(port1 + 1);
1131 porta = -portl;
1132 }
1133 else if (*port1 == '+') { /* positive offset */
Willy Tarreau7f96a842020-09-15 11:12:44 +02001134 if (!(opts & PA_O_PORT_OFS)) {
1135 memprintf(err, "port offset not permitted here in '%s'", str);
1136 goto out;
1137 }
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001138 porth = atoi(port1 + 1);
1139 porta = porth;
1140 }
1141 else if (*port1) { /* other any unexpected char */
Willy Tarreaudad36a32013-03-11 01:20:04 +01001142 memprintf(err, "invalid character '%c' in port number '%s' in '%s'\n", *port1, port1, str);
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001143 goto out;
1144 }
Willy Tarreau7f96a842020-09-15 11:12:44 +02001145 else if (opts & PA_O_PORT_MAND) {
1146 memprintf(err, "missing port specification in '%s'", str);
1147 goto out;
1148 }
Willy Tarreauceccdd72016-11-02 22:27:10 +01001149
1150 /* first try to parse the IP without resolving. If it fails, it
1151 * tells us we need to keep a copy of the FQDN to resolve later
1152 * and to enable DNS. In this case we can proceed if <fqdn> is
Willy Tarreaucd3a55912020-09-04 15:30:46 +02001153 * set or if PA_O_RESOLVE is set, otherwise it's an error.
Willy Tarreauceccdd72016-11-02 22:27:10 +01001154 */
1155 if (str2ip2(str2, &ss, 0) == NULL) {
Willy Tarreaucd3a55912020-09-04 15:30:46 +02001156 if ((!(opts & PA_O_RESOLVE) && !fqdn) ||
1157 ((opts & PA_O_RESOLVE) && str2ip2(str2, &ss, 1) == NULL)) {
Willy Tarreauceccdd72016-11-02 22:27:10 +01001158 memprintf(err, "invalid address: '%s' in '%s'\n", str2, str);
1159 goto out;
1160 }
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001161
Willy Tarreauceccdd72016-11-02 22:27:10 +01001162 if (fqdn) {
1163 if (str2 != back)
1164 memmove(back, str2, strlen(str2) + 1);
1165 *fqdn = back;
1166 back = NULL;
1167 }
Willy Tarreau72b8c1f2015-09-08 15:50:19 +02001168 }
Willy Tarreauceccdd72016-11-02 22:27:10 +01001169 set_host_port(&ss, porta);
Willy Tarreaue4c58c82013-03-06 15:28:17 +01001170 }
Willy Tarreaufab5a432011-03-04 15:31:53 +01001171
Willy Tarreaue835bd82020-09-16 11:35:47 +02001172 if (ctrl_type == SOCK_STREAM && !(opts & PA_O_STREAM)) {
1173 memprintf(err, "stream-type socket not acceptable in '%s'\n", str);
1174 goto out;
1175 }
1176 else if (ctrl_type == SOCK_DGRAM && !(opts & PA_O_DGRAM)) {
1177 memprintf(err, "dgram-type socket not acceptable in '%s'\n", str);
1178 goto out;
1179 }
1180
Willy Tarreau65ec4e32020-09-16 19:17:08 +02001181 if (proto || (opts & PA_O_CONNECT)) {
Willy Tarreau5fc93282020-09-16 18:25:03 +02001182 /* Note: if the caller asks for a proto, we must find one,
1183 * except if we return with an fqdn that will resolve later,
1184 * in which case the address is not known yet (this is only
1185 * for servers actually).
1186 */
Willy Tarreaub2ffc992020-09-16 21:37:31 +02001187 new_proto = protocol_lookup(ss.ss_family,
Willy Tarreauaf9609b2020-09-16 22:04:46 +02001188 sock_type == SOCK_DGRAM,
1189 ctrl_type == SOCK_DGRAM);
Willy Tarreaub2ffc992020-09-16 21:37:31 +02001190
Willy Tarreau5fc93282020-09-16 18:25:03 +02001191 if (!new_proto && (!fqdn || !*fqdn)) {
1192 memprintf(err, "unsupported protocol family %d for address '%s'", ss.ss_family, str);
1193 goto out;
1194 }
Willy Tarreau65ec4e32020-09-16 19:17:08 +02001195
1196 if ((opts & PA_O_CONNECT) && new_proto && !new_proto->connect) {
1197 memprintf(err, "connect() not supported for this protocol family %d used by address '%s'", ss.ss_family, str);
1198 goto out;
1199 }
Willy Tarreau5fc93282020-09-16 18:25:03 +02001200 }
1201
Willy Tarreauc120c8d2013-03-10 19:27:44 +01001202 ret = &ss;
Willy Tarreaud5191e72010-02-09 20:50:45 +01001203 out:
Willy Tarreau48ef4c92017-01-06 18:32:38 +01001204 if (port)
1205 *port = porta;
Willy Tarreaud4448bc2013-02-20 15:55:15 +01001206 if (low)
1207 *low = portl;
1208 if (high)
1209 *high = porth;
Willy Tarreaua93e5c72020-09-15 14:01:16 +02001210 if (fd)
1211 *fd = new_fd;
Willy Tarreau5fc93282020-09-16 18:25:03 +02001212 if (proto)
1213 *proto = new_proto;
Willy Tarreau24709282013-03-10 21:32:12 +01001214 free(back);
Willy Tarreaud5191e72010-02-09 20:50:45 +01001215 return ret;
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02001216}
1217
Thayne McCombs92149f92020-11-20 01:28:26 -07001218/* converts <addr> and <port> into a string representation of the address and port. This is sort
1219 * of an inverse of str2sa_range, with some restrictions. The supported families are AF_INET,
1220 * AF_INET6, AF_UNIX, and AF_CUST_SOCKPAIR. If the family is unsopported NULL is returned.
1221 * If map_ports is true, then the sign of the port is included in the output, to indicate it is
1222 * relative to the incoming port. AF_INET and AF_INET6 will be in the form "<addr>:<port>".
1223 * AF_UNIX will either be just the path (if using a pathname) or "abns@<path>" if it is abstract.
1224 * AF_CUST_SOCKPAIR will be of the form "sockpair@<fd>".
1225 *
1226 * The returned char* is allocated, and it is the responsibility of the caller to free it.
1227 */
1228char * sa2str(const struct sockaddr_storage *addr, int port, int map_ports)
1229{
1230 char buffer[INET6_ADDRSTRLEN];
1231 char *out = NULL;
1232 const void *ptr;
1233 const char *path;
1234
1235 switch (addr->ss_family) {
1236 case AF_INET:
1237 ptr = &((struct sockaddr_in *)addr)->sin_addr;
1238 break;
1239 case AF_INET6:
1240 ptr = &((struct sockaddr_in6 *)addr)->sin6_addr;
1241 break;
1242 case AF_UNIX:
1243 path = ((struct sockaddr_un *)addr)->sun_path;
1244 if (path[0] == '\0') {
1245 const int max_length = sizeof(struct sockaddr_un) - offsetof(struct sockaddr_un, sun_path) - 1;
1246 return memprintf(&out, "abns@%.*s", max_length, path+1);
1247 } else {
1248 return strdup(path);
1249 }
1250 case AF_CUST_SOCKPAIR:
1251 return memprintf(&out, "sockpair@%d", ((struct sockaddr_in *)addr)->sin_addr.s_addr);
1252 default:
1253 return NULL;
1254 }
1255 inet_ntop(addr->ss_family, ptr, buffer, get_addr_len(addr));
1256 if (map_ports)
1257 return memprintf(&out, "%s:%+d", buffer, port);
1258 else
1259 return memprintf(&out, "%s:%d", buffer, port);
1260}
1261
1262
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001263/* converts <str> to a struct in_addr containing a network mask. It can be
1264 * passed in dotted form (255.255.255.0) or in CIDR form (24). It returns 1
Jarno Huuskonen577d5ac2017-05-21 17:32:21 +03001265 * if the conversion succeeds otherwise zero.
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001266 */
1267int str2mask(const char *str, struct in_addr *mask)
1268{
1269 if (strchr(str, '.') != NULL) { /* dotted notation */
1270 if (!inet_pton(AF_INET, str, mask))
1271 return 0;
1272 }
1273 else { /* mask length */
1274 char *err;
1275 unsigned long len = strtol(str, &err, 10);
1276
1277 if (!*str || (err && *err) || (unsigned)len > 32)
1278 return 0;
Tim Duesterhus8575f722018-01-25 16:24:48 +01001279
1280 len2mask4(len, mask);
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001281 }
1282 return 1;
1283}
1284
Tim Duesterhus47185172018-01-25 16:24:49 +01001285/* converts <str> to a struct in6_addr containing a network mask. It can be
Tim Duesterhus5e642862018-02-20 17:02:18 +01001286 * passed in quadruplet form (ffff:ffff::) or in CIDR form (64). It returns 1
Tim Duesterhus47185172018-01-25 16:24:49 +01001287 * if the conversion succeeds otherwise zero.
1288 */
1289int str2mask6(const char *str, struct in6_addr *mask)
1290{
1291 if (strchr(str, ':') != NULL) { /* quadruplet notation */
1292 if (!inet_pton(AF_INET6, str, mask))
1293 return 0;
1294 }
1295 else { /* mask length */
1296 char *err;
1297 unsigned long len = strtol(str, &err, 10);
1298
1299 if (!*str || (err && *err) || (unsigned)len > 128)
1300 return 0;
1301
1302 len2mask6(len, mask);
1303 }
1304 return 1;
1305}
1306
Thierry FOURNIERb0504632013-12-14 15:39:02 +01001307/* convert <cidr> to struct in_addr <mask>. It returns 1 if the conversion
1308 * succeeds otherwise zero.
1309 */
1310int cidr2dotted(int cidr, struct in_addr *mask) {
1311
1312 if (cidr < 0 || cidr > 32)
1313 return 0;
1314
1315 mask->s_addr = cidr ? htonl(~0UL << (32 - cidr)) : 0;
1316 return 1;
1317}
1318
Thierry Fournier70473a52016-02-17 17:12:14 +01001319/* Convert mask from bit length form to in_addr form.
1320 * This function never fails.
1321 */
1322void len2mask4(int len, struct in_addr *addr)
1323{
1324 if (len >= 32) {
1325 addr->s_addr = 0xffffffff;
1326 return;
1327 }
1328 if (len <= 0) {
1329 addr->s_addr = 0x00000000;
1330 return;
1331 }
1332 addr->s_addr = 0xffffffff << (32 - len);
1333 addr->s_addr = htonl(addr->s_addr);
1334}
1335
1336/* Convert mask from bit length form to in6_addr form.
1337 * This function never fails.
1338 */
1339void len2mask6(int len, struct in6_addr *addr)
1340{
1341 len2mask4(len, (struct in_addr *)&addr->s6_addr[0]); /* msb */
1342 len -= 32;
1343 len2mask4(len, (struct in_addr *)&addr->s6_addr[4]);
1344 len -= 32;
1345 len2mask4(len, (struct in_addr *)&addr->s6_addr[8]);
1346 len -= 32;
1347 len2mask4(len, (struct in_addr *)&addr->s6_addr[12]); /* lsb */
1348}
1349
Willy Tarreauc6f4ce82009-06-10 11:09:37 +02001350/*
Willy Tarreaud077a8e2007-05-08 18:28:09 +02001351 * converts <str> to two struct in_addr* which must be pre-allocated.
Willy Tarreaubaaee002006-06-26 02:48:02 +02001352 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001353 * is optional and either in the dotted or CIDR notation.
Willy Tarreaubaaee002006-06-26 02:48:02 +02001354 * Note: "addr" can also be a hostname. Returns 1 if OK, 0 if error.
1355 */
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001356int str2net(const char *str, int resolve, struct in_addr *addr, struct in_addr *mask)
Willy Tarreaubaaee002006-06-26 02:48:02 +02001357{
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001358 __label__ out_free, out_err;
1359 char *c, *s;
1360 int ret_val;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001361
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001362 s = strdup(str);
1363 if (!s)
1364 return 0;
1365
Willy Tarreaubaaee002006-06-26 02:48:02 +02001366 memset(mask, 0, sizeof(*mask));
1367 memset(addr, 0, sizeof(*addr));
Willy Tarreaubaaee002006-06-26 02:48:02 +02001368
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001369 if ((c = strrchr(s, '/')) != NULL) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001370 *c++ = '\0';
1371 /* c points to the mask */
Willy Tarreau2937c0d2010-01-26 17:36:17 +01001372 if (!str2mask(c, mask))
1373 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001374 }
1375 else {
Willy Tarreauebd61602006-12-30 11:54:15 +01001376 mask->s_addr = ~0U;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001377 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001378 if (!inet_pton(AF_INET, s, addr)) {
Willy Tarreaubaaee002006-06-26 02:48:02 +02001379 struct hostent *he;
1380
Thierry FOURNIERfc7ac7b2014-02-11 15:23:04 +01001381 if (!resolve)
1382 goto out_err;
1383
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001384 if ((he = gethostbyname(s)) == NULL) {
1385 goto out_err;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001386 }
1387 else
1388 *addr = *(struct in_addr *) *(he->h_addr_list);
1389 }
Willy Tarreau8aeae4a2007-06-17 11:42:08 +02001390
1391 ret_val = 1;
1392 out_free:
1393 free(s);
1394 return ret_val;
1395 out_err:
1396 ret_val = 0;
1397 goto out_free;
Willy Tarreaubaaee002006-06-26 02:48:02 +02001398}
1399
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001400
1401/*
Willy Tarreau6d20e282012-04-27 22:49:47 +02001402 * converts <str> to two struct in6_addr* which must be pre-allocated.
1403 * The format is "addr[/mask]", where "addr" cannot be empty, and mask
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001404 * is an optional number of bits (128 being the default).
Willy Tarreau6d20e282012-04-27 22:49:47 +02001405 * Returns 1 if OK, 0 if error.
1406 */
1407int str62net(const char *str, struct in6_addr *addr, unsigned char *mask)
1408{
1409 char *c, *s;
1410 int ret_val = 0;
1411 char *err;
1412 unsigned long len = 128;
1413
1414 s = strdup(str);
1415 if (!s)
1416 return 0;
1417
1418 memset(mask, 0, sizeof(*mask));
1419 memset(addr, 0, sizeof(*addr));
1420
1421 if ((c = strrchr(s, '/')) != NULL) {
1422 *c++ = '\0'; /* c points to the mask */
1423 if (!*c)
1424 goto out_free;
1425
1426 len = strtoul(c, &err, 10);
1427 if ((err && *err) || (unsigned)len > 128)
1428 goto out_free;
1429 }
1430 *mask = len; /* OK we have a valid mask in <len> */
1431
1432 if (!inet_pton(AF_INET6, s, addr))
1433 goto out_free;
1434
1435 ret_val = 1;
1436 out_free:
1437 free(s);
1438 return ret_val;
1439}
1440
1441
1442/*
David du Colombier6f5ccb12011-03-10 22:26:24 +01001443 * Parse IPv4 address found in url.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001444 */
David du Colombier6f5ccb12011-03-10 22:26:24 +01001445int url2ipv4(const char *addr, struct in_addr *dst)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001446{
1447 int saw_digit, octets, ch;
1448 u_char tmp[4], *tp;
1449 const char *cp = addr;
1450
1451 saw_digit = 0;
1452 octets = 0;
1453 *(tp = tmp) = 0;
1454
1455 while (*addr) {
1456 unsigned char digit = (ch = *addr++) - '0';
1457 if (digit > 9 && ch != '.')
1458 break;
1459 if (digit <= 9) {
1460 u_int new = *tp * 10 + digit;
1461 if (new > 255)
1462 return 0;
1463 *tp = new;
1464 if (!saw_digit) {
1465 if (++octets > 4)
1466 return 0;
1467 saw_digit = 1;
1468 }
1469 } else if (ch == '.' && saw_digit) {
1470 if (octets == 4)
1471 return 0;
1472 *++tp = 0;
1473 saw_digit = 0;
1474 } else
1475 return 0;
1476 }
1477
1478 if (octets < 4)
1479 return 0;
1480
1481 memcpy(&dst->s_addr, tmp, 4);
1482 return addr-cp-1;
1483}
1484
1485/*
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001486 * Resolve destination server from URL. Convert <str> to a sockaddr_storage.
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05001487 * <out> contain the code of the detected scheme, the start and length of
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001488 * the hostname. Actually only http and https are supported. <out> can be NULL.
1489 * This function returns the consumed length. It is useful if you parse complete
1490 * url like http://host:port/path, because the consumed length corresponds to
1491 * the first character of the path. If the conversion fails, it returns -1.
1492 *
1493 * This function tries to resolve the DNS name if haproxy is in starting mode.
1494 * So, this function may be used during the configuration parsing.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001495 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001496int url2sa(const char *url, int ulen, struct sockaddr_storage *addr, struct split_url *out)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001497{
1498 const char *curr = url, *cp = url;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001499 const char *end;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001500 int ret, url_code = 0;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001501 unsigned long long int http_code = 0;
1502 int default_port;
1503 struct hostent *he;
1504 char *p;
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001505
1506 /* Firstly, try to find :// pattern */
1507 while (curr < url+ulen && url_code != 0x3a2f2f) {
1508 url_code = ((url_code & 0xffff) << 8);
1509 url_code += (unsigned char)*curr++;
1510 }
1511
1512 /* Secondly, if :// pattern is found, verify parsed stuff
1513 * before pattern is matching our http pattern.
1514 * If so parse ip address and port in uri.
1515 *
1516 * WARNING: Current code doesn't support dynamic async dns resolver.
1517 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001518 if (url_code != 0x3a2f2f)
1519 return -1;
1520
1521 /* Copy scheme, and utrn to lower case. */
1522 while (cp < curr - 3)
1523 http_code = (http_code << 8) + *cp++;
1524 http_code |= 0x2020202020202020ULL; /* Turn everything to lower case */
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001525
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001526 /* HTTP or HTTPS url matching */
1527 if (http_code == 0x2020202068747470ULL) {
1528 default_port = 80;
1529 if (out)
1530 out->scheme = SCH_HTTP;
1531 }
1532 else if (http_code == 0x2020206874747073ULL) {
1533 default_port = 443;
1534 if (out)
1535 out->scheme = SCH_HTTPS;
1536 }
1537 else
1538 return -1;
1539
1540 /* If the next char is '[', the host address is IPv6. */
1541 if (*curr == '[') {
1542 curr++;
1543
1544 /* Check trash size */
1545 if (trash.size < ulen)
1546 return -1;
1547
1548 /* Look for ']' and copy the address in a trash buffer. */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001549 p = trash.area;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001550 for (end = curr;
1551 end < url + ulen && *end != ']';
1552 end++, p++)
1553 *p = *end;
1554 if (*end != ']')
1555 return -1;
1556 *p = '\0';
1557
1558 /* Update out. */
1559 if (out) {
1560 out->host = curr;
1561 out->host_len = end - curr;
1562 }
1563
1564 /* Try IPv6 decoding. */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001565 if (!inet_pton(AF_INET6, trash.area, &((struct sockaddr_in6 *)addr)->sin6_addr))
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001566 return -1;
1567 end++;
1568
1569 /* Decode port. */
1570 if (*end == ':') {
1571 end++;
1572 default_port = read_uint(&end, url + ulen);
1573 }
1574 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1575 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1576 return end - url;
1577 }
1578 else {
1579 /* We are looking for IP address. If you want to parse and
1580 * resolve hostname found in url, you can use str2sa_range(), but
1581 * be warned this can slow down global daemon performances
1582 * while handling lagging dns responses.
1583 */
1584 ret = url2ipv4(curr, &((struct sockaddr_in *)addr)->sin_addr);
1585 if (ret) {
1586 /* Update out. */
1587 if (out) {
1588 out->host = curr;
1589 out->host_len = ret;
1590 }
1591
1592 curr += ret;
1593
1594 /* Decode port. */
1595 if (*curr == ':') {
1596 curr++;
1597 default_port = read_uint(&curr, url + ulen);
1598 }
1599 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1600
1601 /* Set family. */
1602 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1603 return curr - url;
1604 }
1605 else if (global.mode & MODE_STARTING) {
1606 /* The IPv4 and IPv6 decoding fails, maybe the url contain name. Try to execute
1607 * synchronous DNS request only if HAProxy is in the start state.
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001608 */
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001609
1610 /* look for : or / or end */
1611 for (end = curr;
1612 end < url + ulen && *end != '/' && *end != ':';
1613 end++);
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001614 memcpy(trash.area, curr, end - curr);
1615 trash.area[end - curr] = '\0';
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001616
1617 /* try to resolve an IPv4/IPv6 hostname */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001618 he = gethostbyname(trash.area);
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001619 if (!he)
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001620 return -1;
Thierry FOURNIER9f95e402014-03-21 14:51:46 +01001621
1622 /* Update out. */
1623 if (out) {
1624 out->host = curr;
1625 out->host_len = end - curr;
1626 }
1627
1628 /* Decode port. */
1629 if (*end == ':') {
1630 end++;
1631 default_port = read_uint(&end, url + ulen);
1632 }
1633
1634 /* Copy IP address, set port and family. */
1635 switch (he->h_addrtype) {
1636 case AF_INET:
1637 ((struct sockaddr_in *)addr)->sin_addr = *(struct in_addr *) *(he->h_addr_list);
1638 ((struct sockaddr_in *)addr)->sin_port = htons(default_port);
1639 ((struct sockaddr_in *)addr)->sin_family = AF_INET;
1640 return end - url;
1641
1642 case AF_INET6:
1643 ((struct sockaddr_in6 *)addr)->sin6_addr = *(struct in6_addr *) *(he->h_addr_list);
1644 ((struct sockaddr_in6 *)addr)->sin6_port = htons(default_port);
1645 ((struct sockaddr_in6 *)addr)->sin6_family = AF_INET6;
1646 return end - url;
1647 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001648 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001649 }
Alexandre Cassen5eb1a902007-11-29 15:43:32 +01001650 return -1;
1651}
1652
Willy Tarreau631f01c2011-09-05 00:36:48 +02001653/* Tries to convert a sockaddr_storage address to text form. Upon success, the
1654 * address family is returned so that it's easy for the caller to adapt to the
1655 * output format. Zero is returned if the address family is not supported. -1
1656 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1657 * supported.
1658 */
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001659int addr_to_str(const struct sockaddr_storage *addr, char *str, int size)
Willy Tarreau631f01c2011-09-05 00:36:48 +02001660{
1661
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001662 const void *ptr;
Willy Tarreau631f01c2011-09-05 00:36:48 +02001663
1664 if (size < 5)
1665 return 0;
1666 *str = '\0';
1667
1668 switch (addr->ss_family) {
1669 case AF_INET:
1670 ptr = &((struct sockaddr_in *)addr)->sin_addr;
1671 break;
1672 case AF_INET6:
1673 ptr = &((struct sockaddr_in6 *)addr)->sin6_addr;
1674 break;
1675 case AF_UNIX:
1676 memcpy(str, "unix", 5);
1677 return addr->ss_family;
1678 default:
1679 return 0;
1680 }
1681
1682 if (inet_ntop(addr->ss_family, ptr, str, size))
1683 return addr->ss_family;
1684
1685 /* failed */
1686 return -1;
1687}
1688
Simon Horman75ab8bd2014-06-16 09:39:41 +09001689/* Tries to convert a sockaddr_storage port to text form. Upon success, the
1690 * address family is returned so that it's easy for the caller to adapt to the
1691 * output format. Zero is returned if the address family is not supported. -1
1692 * is returned upon error, with errno set. AF_INET, AF_INET6 and AF_UNIX are
1693 * supported.
1694 */
Willy Tarreaud5ec4bf2019-04-25 17:48:16 +02001695int port_to_str(const struct sockaddr_storage *addr, char *str, int size)
Simon Horman75ab8bd2014-06-16 09:39:41 +09001696{
1697
1698 uint16_t port;
1699
1700
Willy Tarreaud7dad1b2017-01-06 16:46:22 +01001701 if (size < 6)
Simon Horman75ab8bd2014-06-16 09:39:41 +09001702 return 0;
1703 *str = '\0';
1704
1705 switch (addr->ss_family) {
1706 case AF_INET:
1707 port = ((struct sockaddr_in *)addr)->sin_port;
1708 break;
1709 case AF_INET6:
1710 port = ((struct sockaddr_in6 *)addr)->sin6_port;
1711 break;
1712 case AF_UNIX:
1713 memcpy(str, "unix", 5);
1714 return addr->ss_family;
1715 default:
1716 return 0;
1717 }
1718
1719 snprintf(str, size, "%u", ntohs(port));
1720 return addr->ss_family;
1721}
1722
Willy Tarreau16e01562016-08-09 16:46:18 +02001723/* check if the given address is local to the system or not. It will return
1724 * -1 when it's not possible to know, 0 when the address is not local, 1 when
1725 * it is. We don't want to iterate over all interfaces for this (and it is not
1726 * portable). So instead we try to bind in UDP to this address on a free non
1727 * privileged port and to connect to the same address, port 0 (connect doesn't
1728 * care). If it succeeds, we own the address. Note that non-inet addresses are
1729 * considered local since they're most likely AF_UNIX.
1730 */
1731int addr_is_local(const struct netns_entry *ns,
1732 const struct sockaddr_storage *orig)
1733{
1734 struct sockaddr_storage addr;
1735 int result;
1736 int fd;
1737
1738 if (!is_inet_addr(orig))
1739 return 1;
1740
1741 memcpy(&addr, orig, sizeof(addr));
1742 set_host_port(&addr, 0);
1743
1744 fd = my_socketat(ns, addr.ss_family, SOCK_DGRAM, IPPROTO_UDP);
1745 if (fd < 0)
1746 return -1;
1747
1748 result = -1;
1749 if (bind(fd, (struct sockaddr *)&addr, get_addr_len(&addr)) == 0) {
1750 if (connect(fd, (struct sockaddr *)&addr, get_addr_len(&addr)) == -1)
1751 result = 0; // fail, non-local address
1752 else
1753 result = 1; // success, local address
1754 }
1755 else {
1756 if (errno == EADDRNOTAVAIL)
1757 result = 0; // definitely not local :-)
1758 }
1759 close(fd);
1760
1761 return result;
1762}
1763
Willy Tarreaubaaee002006-06-26 02:48:02 +02001764/* will try to encode the string <string> replacing all characters tagged in
1765 * <map> with the hexadecimal representation of their ASCII-code (2 digits)
1766 * prefixed by <escape>, and will store the result between <start> (included)
1767 * and <stop> (excluded), and will always terminate the string with a '\0'
1768 * before <stop>. The position of the '\0' is returned if the conversion
1769 * completes. If bytes are missing between <start> and <stop>, then the
1770 * conversion will be incomplete and truncated. If <stop> <= <start>, the '\0'
1771 * cannot even be stored so we return <start> without writing the 0.
1772 * The input string must also be zero-terminated.
1773 */
1774const char hextab[16] = "0123456789ABCDEF";
1775char *encode_string(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001776 const char escape, const long *map,
Willy Tarreaubaaee002006-06-26 02:48:02 +02001777 const char *string)
1778{
1779 if (start < stop) {
1780 stop--; /* reserve one byte for the final '\0' */
1781 while (start < stop && *string != '\0') {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001782 if (!ha_bit_test((unsigned char)(*string), map))
Willy Tarreaubaaee002006-06-26 02:48:02 +02001783 *start++ = *string;
1784 else {
1785 if (start + 3 >= stop)
1786 break;
1787 *start++ = escape;
1788 *start++ = hextab[(*string >> 4) & 15];
1789 *start++ = hextab[*string & 15];
1790 }
1791 string++;
1792 }
1793 *start = '\0';
1794 }
1795 return start;
1796}
1797
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001798/*
1799 * Same behavior as encode_string() above, except that it encodes chunk
1800 * <chunk> instead of a string.
1801 */
1802char *encode_chunk(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001803 const char escape, const long *map,
Willy Tarreau83061a82018-07-13 11:56:34 +02001804 const struct buffer *chunk)
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001805{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001806 char *str = chunk->area;
1807 char *end = chunk->area + chunk->data;
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001808
1809 if (start < stop) {
1810 stop--; /* reserve one byte for the final '\0' */
1811 while (start < stop && str < end) {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001812 if (!ha_bit_test((unsigned char)(*str), map))
Thierry FOURNIERe059ec92014-03-17 12:01:13 +01001813 *start++ = *str;
1814 else {
1815 if (start + 3 >= stop)
1816 break;
1817 *start++ = escape;
1818 *start++ = hextab[(*str >> 4) & 15];
1819 *start++ = hextab[*str & 15];
1820 }
1821 str++;
1822 }
1823 *start = '\0';
1824 }
1825 return start;
1826}
1827
Dragan Dosen0edd1092016-02-12 13:23:02 +01001828/*
1829 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001830 * character. The input <string> must be zero-terminated. The result will
1831 * be stored between <start> (included) and <stop> (excluded). This
1832 * function will always try to terminate the resulting string with a '\0'
1833 * before <stop>, and will return its position if the conversion
1834 * completes.
1835 */
1836char *escape_string(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001837 const char escape, const long *map,
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001838 const char *string)
1839{
1840 if (start < stop) {
1841 stop--; /* reserve one byte for the final '\0' */
1842 while (start < stop && *string != '\0') {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001843 if (!ha_bit_test((unsigned char)(*string), map))
Dragan Dosen1a5d0602016-07-22 16:00:31 +02001844 *start++ = *string;
1845 else {
1846 if (start + 2 >= stop)
1847 break;
1848 *start++ = escape;
1849 *start++ = *string;
1850 }
1851 string++;
1852 }
1853 *start = '\0';
1854 }
1855 return start;
1856}
1857
1858/*
1859 * Tries to prefix characters tagged in the <map> with the <escape>
Dragan Dosen0edd1092016-02-12 13:23:02 +01001860 * character. <chunk> contains the input to be escaped. The result will be
1861 * stored between <start> (included) and <stop> (excluded). The function
1862 * will always try to terminate the resulting string with a '\0' before
1863 * <stop>, and will return its position if the conversion completes.
1864 */
1865char *escape_chunk(char *start, char *stop,
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001866 const char escape, const long *map,
Willy Tarreau83061a82018-07-13 11:56:34 +02001867 const struct buffer *chunk)
Dragan Dosen0edd1092016-02-12 13:23:02 +01001868{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001869 char *str = chunk->area;
1870 char *end = chunk->area + chunk->data;
Dragan Dosen0edd1092016-02-12 13:23:02 +01001871
1872 if (start < stop) {
1873 stop--; /* reserve one byte for the final '\0' */
1874 while (start < stop && str < end) {
Willy Tarreau1bfd6022019-06-07 11:10:07 +02001875 if (!ha_bit_test((unsigned char)(*str), map))
Dragan Dosen0edd1092016-02-12 13:23:02 +01001876 *start++ = *str;
1877 else {
1878 if (start + 2 >= stop)
1879 break;
1880 *start++ = escape;
1881 *start++ = *str;
1882 }
1883 str++;
1884 }
1885 *start = '\0';
1886 }
1887 return start;
1888}
1889
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001890/* Check a string for using it in a CSV output format. If the string contains
1891 * one of the following four char <">, <,>, CR or LF, the string is
1892 * encapsulated between <"> and the <"> are escaped by a <""> sequence.
1893 * <str> is the input string to be escaped. The function assumes that
1894 * the input string is null-terminated.
1895 *
1896 * If <quote> is 0, the result is returned escaped but without double quote.
Willy Tarreau898529b2016-01-06 18:07:04 +01001897 * It is useful if the escaped string is used between double quotes in the
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001898 * format.
1899 *
Willy Tarreau898529b2016-01-06 18:07:04 +01001900 * printf("..., \"%s\", ...\r\n", csv_enc(str, 0, &trash));
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001901 *
Willy Tarreaub631c292016-01-08 10:04:08 +01001902 * If <quote> is 1, the converter puts the quotes only if any reserved character
1903 * is present. If <quote> is 2, the converter always puts the quotes.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001904 *
Willy Tarreau83061a82018-07-13 11:56:34 +02001905 * <output> is a struct buffer used for storing the output string.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001906 *
Willy Tarreau898529b2016-01-06 18:07:04 +01001907 * The function returns the converted string on its output. If an error
1908 * occurs, the function returns an empty string. This type of output is useful
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001909 * for using the function directly as printf() argument.
1910 *
1911 * If the output buffer is too short to contain the input string, the result
1912 * is truncated.
Willy Tarreau898529b2016-01-06 18:07:04 +01001913 *
Willy Tarreaub631c292016-01-08 10:04:08 +01001914 * This function appends the encoding to the existing output chunk, and it
1915 * guarantees that it starts immediately at the first available character of
1916 * the chunk. Please use csv_enc() instead if you want to replace the output
1917 * chunk.
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001918 */
Willy Tarreau83061a82018-07-13 11:56:34 +02001919const char *csv_enc_append(const char *str, int quote, struct buffer *output)
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001920{
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001921 char *end = output->area + output->size;
1922 char *out = output->area + output->data;
Willy Tarreau898529b2016-01-06 18:07:04 +01001923 char *ptr = out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001924
Willy Tarreaub631c292016-01-08 10:04:08 +01001925 if (quote == 1) {
1926 /* automatic quoting: first verify if we'll have to quote the string */
1927 if (!strpbrk(str, "\n\r,\""))
1928 quote = 0;
1929 }
1930
1931 if (quote)
1932 *ptr++ = '"';
1933
Willy Tarreau898529b2016-01-06 18:07:04 +01001934 while (*str && ptr < end - 2) { /* -2 for reserving space for <"> and \0. */
1935 *ptr = *str;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001936 if (*str == '"') {
Willy Tarreau898529b2016-01-06 18:07:04 +01001937 ptr++;
1938 if (ptr >= end - 2) {
1939 ptr--;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001940 break;
1941 }
Willy Tarreau898529b2016-01-06 18:07:04 +01001942 *ptr = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001943 }
Willy Tarreau898529b2016-01-06 18:07:04 +01001944 ptr++;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001945 str++;
1946 }
1947
Willy Tarreaub631c292016-01-08 10:04:08 +01001948 if (quote)
1949 *ptr++ = '"';
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001950
Willy Tarreau898529b2016-01-06 18:07:04 +01001951 *ptr = '\0';
Willy Tarreau843b7cb2018-07-13 10:54:26 +02001952 output->data = ptr - output->area;
Willy Tarreau898529b2016-01-06 18:07:04 +01001953 return out;
Thierry FOURNIERddea6262015-05-28 16:00:28 +02001954}
1955
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001956/* Decode an URL-encoded string in-place. The resulting string might
1957 * be shorter. If some forbidden characters are found, the conversion is
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001958 * aborted, the string is truncated before the issue and a negative value is
1959 * returned, otherwise the operation returns the length of the decoded string.
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001960 * If the 'in_form' argument is non-nul the string is assumed to be part of
1961 * an "application/x-www-form-urlencoded" encoded string, and the '+' will be
1962 * turned to a space. If it's zero, this will only be done after a question
1963 * mark ('?').
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001964 */
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001965int url_decode(char *string, int in_form)
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001966{
1967 char *in, *out;
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001968 int ret = -1;
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001969
1970 in = string;
1971 out = string;
1972 while (*in) {
1973 switch (*in) {
1974 case '+' :
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001975 *out++ = in_form ? ' ' : *in;
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001976 break;
1977 case '%' :
1978 if (!ishex(in[1]) || !ishex(in[2]))
1979 goto end;
1980 *out++ = (hex2i(in[1]) << 4) + hex2i(in[2]);
1981 in += 2;
1982 break;
Willy Tarreau62ba9ba2020-04-23 17:54:47 +02001983 case '?':
1984 in_form = 1;
1985 /* fall through */
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001986 default:
1987 *out++ = *in;
1988 break;
1989 }
1990 in++;
1991 }
Thierry FOURNIER5068d962013-10-04 16:27:27 +02001992 ret = out - string; /* success */
Willy Tarreaubf9c2fc2011-05-31 18:06:18 +02001993 end:
1994 *out = 0;
1995 return ret;
1996}
Willy Tarreaubaaee002006-06-26 02:48:02 +02001997
Willy Tarreau6911fa42007-03-04 18:06:08 +01001998unsigned int str2ui(const char *s)
1999{
2000 return __str2ui(s);
2001}
2002
2003unsigned int str2uic(const char *s)
2004{
2005 return __str2uic(s);
2006}
2007
2008unsigned int strl2ui(const char *s, int len)
2009{
2010 return __strl2ui(s, len);
2011}
2012
2013unsigned int strl2uic(const char *s, int len)
2014{
2015 return __strl2uic(s, len);
2016}
2017
Willy Tarreau4ec83cd2010-10-15 23:19:55 +02002018unsigned int read_uint(const char **s, const char *end)
2019{
2020 return __read_uint(s, end);
2021}
2022
Thierry FOURNIER763a5d82015-07-06 23:09:52 +02002023/* This function reads an unsigned integer from the string pointed to by <s> and
2024 * returns it. The <s> pointer is adjusted to point to the first unread char. The
2025 * function automatically stops at <end>. If the number overflows, the 2^64-1
2026 * value is returned.
2027 */
2028unsigned long long int read_uint64(const char **s, const char *end)
2029{
2030 const char *ptr = *s;
2031 unsigned long long int i = 0, tmp;
2032 unsigned int j;
2033
2034 while (ptr < end) {
2035
2036 /* read next char */
2037 j = *ptr - '0';
2038 if (j > 9)
2039 goto read_uint64_end;
2040
2041 /* add char to the number and check overflow. */
2042 tmp = i * 10;
2043 if (tmp / 10 != i) {
2044 i = ULLONG_MAX;
2045 goto read_uint64_eat;
2046 }
2047 if (ULLONG_MAX - tmp < j) {
2048 i = ULLONG_MAX;
2049 goto read_uint64_eat;
2050 }
2051 i = tmp + j;
2052 ptr++;
2053 }
2054read_uint64_eat:
2055 /* eat each numeric char */
2056 while (ptr < end) {
2057 if ((unsigned int)(*ptr - '0') > 9)
2058 break;
2059 ptr++;
2060 }
2061read_uint64_end:
2062 *s = ptr;
2063 return i;
2064}
2065
2066/* This function reads an integer from the string pointed to by <s> and returns
2067 * it. The <s> pointer is adjusted to point to the first unread char. The function
2068 * automatically stops at <end>. Il the number is bigger than 2^63-2, the 2^63-1
2069 * value is returned. If the number is lowest than -2^63-1, the -2^63 value is
2070 * returned.
2071 */
2072long long int read_int64(const char **s, const char *end)
2073{
2074 unsigned long long int i = 0;
2075 int neg = 0;
2076
2077 /* Look for minus char. */
2078 if (**s == '-') {
2079 neg = 1;
2080 (*s)++;
2081 }
2082 else if (**s == '+')
2083 (*s)++;
2084
2085 /* convert as positive number. */
2086 i = read_uint64(s, end);
2087
2088 if (neg) {
2089 if (i > 0x8000000000000000ULL)
2090 return LLONG_MIN;
2091 return -i;
2092 }
2093 if (i > 0x7fffffffffffffffULL)
2094 return LLONG_MAX;
2095 return i;
2096}
2097
Willy Tarreau6911fa42007-03-04 18:06:08 +01002098/* This one is 7 times faster than strtol() on athlon with checks.
2099 * It returns the value of the number composed of all valid digits read,
2100 * and can process negative numbers too.
2101 */
2102int strl2ic(const char *s, int len)
2103{
2104 int i = 0;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002105 int j, k;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002106
2107 if (len > 0) {
2108 if (*s != '-') {
2109 /* positive number */
2110 while (len-- > 0) {
2111 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002112 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002113 if (j > 9)
2114 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002115 i = k + j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002116 }
2117 } else {
2118 /* negative number */
2119 s++;
2120 while (--len > 0) {
2121 j = (*s++) - '0';
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002122 k = i * 10;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002123 if (j > 9)
2124 break;
Willy Tarreau3f0c9762007-10-25 09:42:24 +02002125 i = k - j;
Willy Tarreau6911fa42007-03-04 18:06:08 +01002126 }
2127 }
2128 }
2129 return i;
2130}
2131
2132
2133/* This function reads exactly <len> chars from <s> and converts them to a
2134 * signed integer which it stores into <ret>. It accurately detects any error
2135 * (truncated string, invalid chars, overflows). It is meant to be used in
2136 * applications designed for hostile environments. It returns zero when the
2137 * number has successfully been converted, non-zero otherwise. When an error
2138 * is returned, the <ret> value is left untouched. It is yet 5 to 40 times
2139 * faster than strtol().
2140 */
2141int strl2irc(const char *s, int len, int *ret)
2142{
2143 int i = 0;
2144 int j;
2145
2146 if (!len)
2147 return 1;
2148
2149 if (*s != '-') {
2150 /* positive number */
2151 while (len-- > 0) {
2152 j = (*s++) - '0';
2153 if (j > 9) return 1; /* invalid char */
2154 if (i > INT_MAX / 10) return 1; /* check for multiply overflow */
2155 i = i * 10;
2156 if (i + j < i) return 1; /* check for addition overflow */
2157 i = i + j;
2158 }
2159 } else {
2160 /* negative number */
2161 s++;
2162 while (--len > 0) {
2163 j = (*s++) - '0';
2164 if (j > 9) return 1; /* invalid char */
2165 if (i < INT_MIN / 10) return 1; /* check for multiply overflow */
2166 i = i * 10;
2167 if (i - j > i) return 1; /* check for subtract overflow */
2168 i = i - j;
2169 }
2170 }
2171 *ret = i;
2172 return 0;
2173}
2174
2175
2176/* This function reads exactly <len> chars from <s> and converts them to a
2177 * signed integer which it stores into <ret>. It accurately detects any error
2178 * (truncated string, invalid chars, overflows). It is meant to be used in
2179 * applications designed for hostile environments. It returns zero when the
2180 * number has successfully been converted, non-zero otherwise. When an error
2181 * is returned, the <ret> value is left untouched. It is about 3 times slower
William Dauchy060ffc82021-02-06 20:47:51 +01002182 * than strl2irc().
Willy Tarreau6911fa42007-03-04 18:06:08 +01002183 */
Willy Tarreau6911fa42007-03-04 18:06:08 +01002184
2185int strl2llrc(const char *s, int len, long long *ret)
2186{
2187 long long i = 0;
2188 int j;
2189
2190 if (!len)
2191 return 1;
2192
2193 if (*s != '-') {
2194 /* positive number */
2195 while (len-- > 0) {
2196 j = (*s++) - '0';
2197 if (j > 9) return 1; /* invalid char */
2198 if (i > LLONG_MAX / 10LL) return 1; /* check for multiply overflow */
2199 i = i * 10LL;
2200 if (i + j < i) return 1; /* check for addition overflow */
2201 i = i + j;
2202 }
2203 } else {
2204 /* negative number */
2205 s++;
2206 while (--len > 0) {
2207 j = (*s++) - '0';
2208 if (j > 9) return 1; /* invalid char */
2209 if (i < LLONG_MIN / 10LL) return 1; /* check for multiply overflow */
2210 i = i * 10LL;
2211 if (i - j > i) return 1; /* check for subtract overflow */
2212 i = i - j;
2213 }
2214 }
2215 *ret = i;
2216 return 0;
2217}
2218
Thierry FOURNIER511e9472014-01-23 17:40:34 +01002219/* This function is used with pat_parse_dotted_ver(). It converts a string
2220 * composed by two number separated by a dot. Each part must contain in 16 bits
2221 * because internally they will be represented as a 32-bit quantity stored in
2222 * a 64-bit integer. It returns zero when the number has successfully been
2223 * converted, non-zero otherwise. When an error is returned, the <ret> value
2224 * is left untouched.
2225 *
2226 * "1.3" -> 0x0000000000010003
2227 * "65535.65535" -> 0x00000000ffffffff
2228 */
2229int strl2llrc_dotted(const char *text, int len, long long *ret)
2230{
2231 const char *end = &text[len];
2232 const char *p;
2233 long long major, minor;
2234
2235 /* Look for dot. */
2236 for (p = text; p < end; p++)
2237 if (*p == '.')
2238 break;
2239
2240 /* Convert major. */
2241 if (strl2llrc(text, p - text, &major) != 0)
2242 return 1;
2243
2244 /* Check major. */
2245 if (major >= 65536)
2246 return 1;
2247
2248 /* Convert minor. */
2249 minor = 0;
2250 if (p < end)
2251 if (strl2llrc(p + 1, end - (p + 1), &minor) != 0)
2252 return 1;
2253
2254 /* Check minor. */
2255 if (minor >= 65536)
2256 return 1;
2257
2258 /* Compose value. */
2259 *ret = (major << 16) | (minor & 0xffff);
2260 return 0;
2261}
2262
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002263/* This function parses a time value optionally followed by a unit suffix among
2264 * "d", "h", "m", "s", "ms" or "us". It converts the value into the unit
2265 * expected by the caller. The computation does its best to avoid overflows.
2266 * The value is returned in <ret> if everything is fine, and a NULL is returned
2267 * by the function. In case of error, a pointer to the error is returned and
2268 * <ret> is left untouched. Values are automatically rounded up when needed.
Willy Tarreau9faebe32019-06-07 19:00:37 +02002269 * Values resulting in values larger than or equal to 2^31 after conversion are
2270 * reported as an overflow as value PARSE_TIME_OVER. Non-null values resulting
2271 * in an underflow are reported as an underflow as value PARSE_TIME_UNDER.
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002272 */
2273const char *parse_time_err(const char *text, unsigned *ret, unsigned unit_flags)
2274{
Willy Tarreau9faebe32019-06-07 19:00:37 +02002275 unsigned long long imult, idiv;
2276 unsigned long long omult, odiv;
2277 unsigned long long value, result;
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002278 const char *str = text;
2279
2280 if (!isdigit((unsigned char)*text))
2281 return text;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002282
2283 omult = odiv = 1;
2284
2285 switch (unit_flags & TIME_UNIT_MASK) {
2286 case TIME_UNIT_US: omult = 1000000; break;
2287 case TIME_UNIT_MS: omult = 1000; break;
2288 case TIME_UNIT_S: break;
2289 case TIME_UNIT_MIN: odiv = 60; break;
2290 case TIME_UNIT_HOUR: odiv = 3600; break;
2291 case TIME_UNIT_DAY: odiv = 86400; break;
2292 default: break;
2293 }
2294
2295 value = 0;
2296
2297 while (1) {
2298 unsigned int j;
2299
2300 j = *text - '0';
2301 if (j > 9)
2302 break;
2303 text++;
2304 value *= 10;
2305 value += j;
2306 }
2307
2308 imult = idiv = 1;
2309 switch (*text) {
2310 case '\0': /* no unit = default unit */
2311 imult = omult = idiv = odiv = 1;
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002312 goto end;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002313 case 's': /* second = unscaled unit */
2314 break;
2315 case 'u': /* microsecond : "us" */
2316 if (text[1] == 's') {
2317 idiv = 1000000;
2318 text++;
2319 }
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002320 return text;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002321 case 'm': /* millisecond : "ms" or minute: "m" */
2322 if (text[1] == 's') {
2323 idiv = 1000;
2324 text++;
2325 } else
2326 imult = 60;
2327 break;
2328 case 'h': /* hour : "h" */
2329 imult = 3600;
2330 break;
2331 case 'd': /* day : "d" */
2332 imult = 86400;
2333 break;
2334 default:
2335 return text;
2336 break;
2337 }
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002338 if (*(++text) != '\0') {
2339 ha_warning("unexpected character '%c' after the timer value '%s', only "
2340 "(us=microseconds,ms=milliseconds,s=seconds,m=minutes,h=hours,d=days) are supported."
2341 " This will be reported as an error in next versions.\n", *text, str);
2342 }
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002343
Christopher Fauletc20ad0d2020-12-11 09:23:07 +01002344 end:
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002345 if (omult % idiv == 0) { omult /= idiv; idiv = 1; }
2346 if (idiv % omult == 0) { idiv /= omult; omult = 1; }
2347 if (imult % odiv == 0) { imult /= odiv; odiv = 1; }
2348 if (odiv % imult == 0) { odiv /= imult; imult = 1; }
2349
Willy Tarreau9faebe32019-06-07 19:00:37 +02002350 result = (value * (imult * omult) + (idiv * odiv - 1)) / (idiv * odiv);
2351 if (result >= 0x80000000)
2352 return PARSE_TIME_OVER;
2353 if (!result && value)
2354 return PARSE_TIME_UNDER;
2355 *ret = result;
Willy Tarreaua0d37b62007-12-02 22:00:35 +01002356 return NULL;
2357}
Willy Tarreau6911fa42007-03-04 18:06:08 +01002358
Emeric Brun39132b22010-01-04 14:57:24 +01002359/* this function converts the string starting at <text> to an unsigned int
2360 * stored in <ret>. If an error is detected, the pointer to the unexpected
Joseph Herlant32b83272018-11-15 11:58:28 -08002361 * character is returned. If the conversion is successful, NULL is returned.
Emeric Brun39132b22010-01-04 14:57:24 +01002362 */
2363const char *parse_size_err(const char *text, unsigned *ret) {
2364 unsigned value = 0;
2365
Christopher Faulet82635a02020-12-11 09:30:45 +01002366 if (!isdigit((unsigned char)*text))
2367 return text;
2368
Emeric Brun39132b22010-01-04 14:57:24 +01002369 while (1) {
2370 unsigned int j;
2371
2372 j = *text - '0';
2373 if (j > 9)
2374 break;
2375 if (value > ~0U / 10)
2376 return text;
2377 value *= 10;
2378 if (value > (value + j))
2379 return text;
2380 value += j;
2381 text++;
2382 }
2383
2384 switch (*text) {
2385 case '\0':
2386 break;
2387 case 'K':
2388 case 'k':
2389 if (value > ~0U >> 10)
2390 return text;
2391 value = value << 10;
2392 break;
2393 case 'M':
2394 case 'm':
2395 if (value > ~0U >> 20)
2396 return text;
2397 value = value << 20;
2398 break;
2399 case 'G':
2400 case 'g':
2401 if (value > ~0U >> 30)
2402 return text;
2403 value = value << 30;
2404 break;
2405 default:
2406 return text;
2407 }
2408
Godbach58048a22015-01-28 17:36:16 +08002409 if (*text != '\0' && *++text != '\0')
2410 return text;
2411
Emeric Brun39132b22010-01-04 14:57:24 +01002412 *ret = value;
2413 return NULL;
2414}
2415
Willy Tarreau126d4062013-12-03 17:50:47 +01002416/*
2417 * Parse binary string written in hexadecimal (source) and store the decoded
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002418 * result into binstr and set binstrlen to the length of binstr. Memory for
Willy Tarreau126d4062013-12-03 17:50:47 +01002419 * binstr is allocated by the function. In case of error, returns 0 with an
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002420 * error message in err. In success case, it returns the consumed length.
Willy Tarreau126d4062013-12-03 17:50:47 +01002421 */
2422int parse_binary(const char *source, char **binstr, int *binstrlen, char **err)
2423{
2424 int len;
2425 const char *p = source;
2426 int i,j;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002427 int alloc;
Willy Tarreau126d4062013-12-03 17:50:47 +01002428
2429 len = strlen(source);
2430 if (len % 2) {
2431 memprintf(err, "an even number of hex digit is expected");
2432 return 0;
2433 }
2434
2435 len = len >> 1;
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002436
Willy Tarreau126d4062013-12-03 17:50:47 +01002437 if (!*binstr) {
Tim Duesterhuse52b6e52020-09-12 20:26:43 +02002438 *binstr = calloc(len, sizeof(**binstr));
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002439 if (!*binstr) {
2440 memprintf(err, "out of memory while loading string pattern");
2441 return 0;
2442 }
2443 alloc = 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002444 }
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002445 else {
2446 if (*binstrlen < len) {
Joseph Herlant76dbe782018-11-15 12:01:22 -08002447 memprintf(err, "no space available in the buffer. expect %d, provides %d",
Thierry FOURNIER9645d422013-12-06 19:59:28 +01002448 len, *binstrlen);
2449 return 0;
2450 }
2451 alloc = 0;
2452 }
2453 *binstrlen = len;
Willy Tarreau126d4062013-12-03 17:50:47 +01002454
2455 i = j = 0;
2456 while (j < len) {
2457 if (!ishex(p[i++]))
2458 goto bad_input;
2459 if (!ishex(p[i++]))
2460 goto bad_input;
2461 (*binstr)[j++] = (hex2i(p[i-2]) << 4) + hex2i(p[i-1]);
2462 }
Thierry FOURNIERee330af2014-01-21 11:36:14 +01002463 return len << 1;
Willy Tarreau126d4062013-12-03 17:50:47 +01002464
2465bad_input:
2466 memprintf(err, "an hex digit is expected (found '%c')", p[i-1]);
Andreas Seltenreich93f91c32016-03-03 20:40:37 +01002467 if (alloc) {
2468 free(*binstr);
2469 *binstr = NULL;
2470 }
Willy Tarreau126d4062013-12-03 17:50:47 +01002471 return 0;
2472}
2473
Willy Tarreau946ba592009-05-10 15:41:18 +02002474/* copies at most <n> characters from <src> and always terminates with '\0' */
2475char *my_strndup(const char *src, int n)
2476{
2477 int len = 0;
2478 char *ret;
2479
2480 while (len < n && src[len])
2481 len++;
2482
Vincent Bernat3c2f2f22016-04-03 13:48:42 +02002483 ret = malloc(len + 1);
Willy Tarreau946ba592009-05-10 15:41:18 +02002484 if (!ret)
2485 return ret;
2486 memcpy(ret, src, len);
2487 ret[len] = '\0';
2488 return ret;
2489}
2490
Baptiste Assmannbb77c8e2013-10-06 23:24:13 +02002491/*
2492 * search needle in haystack
2493 * returns the pointer if found, returns NULL otherwise
2494 */
2495const void *my_memmem(const void *haystack, size_t haystacklen, const void *needle, size_t needlelen)
2496{
2497 const void *c = NULL;
2498 unsigned char f;
2499
2500 if ((haystack == NULL) || (needle == NULL) || (haystacklen < needlelen))
2501 return NULL;
2502
2503 f = *(char *)needle;
2504 c = haystack;
2505 while ((c = memchr(c, f, haystacklen - (c - haystack))) != NULL) {
2506 if ((haystacklen - (c - haystack)) < needlelen)
2507 return NULL;
2508
2509 if (memcmp(c, needle, needlelen) == 0)
2510 return c;
2511 ++c;
2512 }
2513 return NULL;
2514}
2515
Ilya Shipitsinc02a23f2020-05-06 00:53:22 +05002516/* get length of the initial segment consisting entirely of bytes in <accept> */
Christopher Faulet5eb96cb2020-04-15 10:23:01 +02002517size_t my_memspn(const void *str, size_t len, const void *accept, size_t acceptlen)
2518{
2519 size_t ret = 0;
2520
2521 while (ret < len && memchr(accept, *((int *)str), acceptlen)) {
2522 str++;
2523 ret++;
2524 }
2525 return ret;
2526}
2527
Ilya Shipitsinc02a23f2020-05-06 00:53:22 +05002528/* get length of the initial segment consisting entirely of bytes not in <rejcet> */
Christopher Faulet5eb96cb2020-04-15 10:23:01 +02002529size_t my_memcspn(const void *str, size_t len, const void *reject, size_t rejectlen)
2530{
2531 size_t ret = 0;
2532
2533 while (ret < len) {
2534 if(memchr(reject, *((int *)str), rejectlen))
2535 return ret;
2536 str++;
2537 ret++;
2538 }
2539 return ret;
2540}
2541
Willy Tarreau482b00d2009-10-04 22:48:42 +02002542/* This function returns the first unused key greater than or equal to <key> in
2543 * ID tree <root>. Zero is returned if no place is found.
2544 */
2545unsigned int get_next_id(struct eb_root *root, unsigned int key)
2546{
2547 struct eb32_node *used;
2548
2549 do {
2550 used = eb32_lookup_ge(root, key);
2551 if (!used || used->key > key)
2552 return key; /* key is available */
2553 key++;
2554 } while (key);
2555 return key;
2556}
2557
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002558/* dump the full tree to <file> in DOT format for debugging purposes. Will
2559 * optionally highlight node <subj> if found, depending on operation <op> :
2560 * 0 : nothing
2561 * >0 : insertion, node/leaf are surrounded in red
2562 * <0 : removal, node/leaf are dashed with no background
2563 * Will optionally add "desc" as a label on the graph if set and non-null.
2564 */
2565void eb32sc_to_file(FILE *file, struct eb_root *root, const struct eb32sc_node *subj, int op, const char *desc)
Willy Tarreaued3cda02017-11-15 15:04:05 +01002566{
2567 struct eb32sc_node *node;
2568 unsigned long scope = -1;
2569
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002570 fprintf(file, "digraph ebtree {\n");
2571
2572 if (desc && *desc) {
2573 fprintf(file,
2574 " fontname=\"fixed\";\n"
2575 " fontsize=8;\n"
2576 " label=\"%s\";\n", desc);
2577 }
2578
Willy Tarreaued3cda02017-11-15 15:04:05 +01002579 fprintf(file,
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002580 " node [fontname=\"fixed\" fontsize=8 shape=\"box\" style=\"filled\" color=\"black\" fillcolor=\"white\"];\n"
2581 " edge [fontname=\"fixed\" fontsize=8 style=\"solid\" color=\"magenta\" dir=\"forward\"];\n"
Willy Tarreaued3cda02017-11-15 15:04:05 +01002582 " \"%lx_n\" [label=\"root\\n%lx\"]\n", (long)eb_root_to_node(root), (long)root
2583 );
2584
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002585 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"L\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002586 (long)eb_root_to_node(root),
2587 (long)eb_root_to_node(eb_clrtag(root->b[0])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002588 eb_gettag(root->b[0]) == EB_LEAF ? 'l' : 'n');
2589
2590 node = eb32sc_first(root, scope);
2591 while (node) {
2592 if (node->node.node_p) {
2593 /* node part is used */
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002594 fprintf(file, " \"%lx_n\" [label=\"%lx\\nkey=%u\\nscope=%lx\\nbit=%d\" fillcolor=\"lightskyblue1\" %s];\n",
2595 (long)node, (long)node, node->key, node->node_s, node->node.bit,
2596 (node == subj) ? (op < 0 ? "color=\"red\" style=\"dashed\"" : op > 0 ? "color=\"red\"" : "") : "");
Willy Tarreaued3cda02017-11-15 15:04:05 +01002597
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002598 fprintf(file, " \"%lx_n\" -> \"%lx_n\" [taillabel=\"%c\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002599 (long)node,
2600 (long)eb_root_to_node(eb_clrtag(node->node.node_p)),
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002601 eb_gettag(node->node.node_p) ? 'R' : 'L');
Willy Tarreaued3cda02017-11-15 15:04:05 +01002602
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002603 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"L\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002604 (long)node,
2605 (long)eb_root_to_node(eb_clrtag(node->node.branches.b[0])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002606 eb_gettag(node->node.branches.b[0]) == EB_LEAF ? 'l' : 'n');
2607
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002608 fprintf(file, " \"%lx_n\" -> \"%lx_%c\" [taillabel=\"R\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002609 (long)node,
2610 (long)eb_root_to_node(eb_clrtag(node->node.branches.b[1])),
Willy Tarreaued3cda02017-11-15 15:04:05 +01002611 eb_gettag(node->node.branches.b[1]) == EB_LEAF ? 'l' : 'n');
2612 }
2613
Willy Tarreau9c1e15d2017-11-15 18:51:29 +01002614 fprintf(file, " \"%lx_l\" [label=\"%lx\\nkey=%u\\nscope=%lx\\npfx=%u\" fillcolor=\"yellow\" %s];\n",
2615 (long)node, (long)node, node->key, node->leaf_s, node->node.pfx,
2616 (node == subj) ? (op < 0 ? "color=\"red\" style=\"dashed\"" : op > 0 ? "color=\"red\"" : "") : "");
Willy Tarreaued3cda02017-11-15 15:04:05 +01002617
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002618 fprintf(file, " \"%lx_l\" -> \"%lx_n\" [taillabel=\"%c\"];\n",
Willy Tarreaued3cda02017-11-15 15:04:05 +01002619 (long)node,
2620 (long)eb_root_to_node(eb_clrtag(node->node.leaf_p)),
Willy Tarreau6c7f4de2017-11-15 17:49:54 +01002621 eb_gettag(node->node.leaf_p) ? 'R' : 'L');
Willy Tarreaued3cda02017-11-15 15:04:05 +01002622 node = eb32sc_next(node, scope);
2623 }
2624 fprintf(file, "}\n");
2625}
2626
Willy Tarreau348238b2010-01-18 15:05:57 +01002627/* This function compares a sample word possibly followed by blanks to another
2628 * clean word. The compare is case-insensitive. 1 is returned if both are equal,
2629 * otherwise zero. This intends to be used when checking HTTP headers for some
2630 * values. Note that it validates a word followed only by blanks but does not
2631 * validate a word followed by blanks then other chars.
2632 */
2633int word_match(const char *sample, int slen, const char *word, int wlen)
2634{
2635 if (slen < wlen)
2636 return 0;
2637
2638 while (wlen) {
2639 char c = *sample ^ *word;
2640 if (c && c != ('A' ^ 'a'))
2641 return 0;
2642 sample++;
2643 word++;
2644 slen--;
2645 wlen--;
2646 }
2647
2648 while (slen) {
2649 if (*sample != ' ' && *sample != '\t')
2650 return 0;
2651 sample++;
2652 slen--;
2653 }
2654 return 1;
2655}
Willy Tarreau482b00d2009-10-04 22:48:42 +02002656
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002657/* Converts any text-formatted IPv4 address to a host-order IPv4 address. It
2658 * is particularly fast because it avoids expensive operations such as
2659 * multiplies, which are optimized away at the end. It requires a properly
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05002660 * formatted address though (3 points).
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002661 */
2662unsigned int inetaddr_host(const char *text)
2663{
2664 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2665 register unsigned int dig100, dig10, dig1;
2666 int s;
2667 const char *p, *d;
2668
2669 dig1 = dig10 = dig100 = ascii_zero;
2670 s = 24;
2671
2672 p = text;
2673 while (1) {
2674 if (((unsigned)(*p - '0')) <= 9) {
2675 p++;
2676 continue;
2677 }
2678
2679 /* here, we have a complete byte between <text> and <p> (exclusive) */
2680 if (p == text)
2681 goto end;
2682
2683 d = p - 1;
2684 dig1 |= (unsigned int)(*d << s);
2685 if (d == text)
2686 goto end;
2687
2688 d--;
2689 dig10 |= (unsigned int)(*d << s);
2690 if (d == text)
2691 goto end;
2692
2693 d--;
2694 dig100 |= (unsigned int)(*d << s);
2695 end:
2696 if (!s || *p != '.')
2697 break;
2698
2699 s -= 8;
2700 text = ++p;
2701 }
2702
2703 dig100 -= ascii_zero;
2704 dig10 -= ascii_zero;
2705 dig1 -= ascii_zero;
2706 return ((dig100 * 10) + dig10) * 10 + dig1;
2707}
2708
2709/*
2710 * Idem except the first unparsed character has to be passed in <stop>.
2711 */
2712unsigned int inetaddr_host_lim(const char *text, const char *stop)
2713{
2714 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2715 register unsigned int dig100, dig10, dig1;
2716 int s;
2717 const char *p, *d;
2718
2719 dig1 = dig10 = dig100 = ascii_zero;
2720 s = 24;
2721
2722 p = text;
2723 while (1) {
2724 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2725 p++;
2726 continue;
2727 }
2728
2729 /* here, we have a complete byte between <text> and <p> (exclusive) */
2730 if (p == text)
2731 goto end;
2732
2733 d = p - 1;
2734 dig1 |= (unsigned int)(*d << s);
2735 if (d == text)
2736 goto end;
2737
2738 d--;
2739 dig10 |= (unsigned int)(*d << s);
2740 if (d == text)
2741 goto end;
2742
2743 d--;
2744 dig100 |= (unsigned int)(*d << s);
2745 end:
2746 if (!s || p == stop || *p != '.')
2747 break;
2748
2749 s -= 8;
2750 text = ++p;
2751 }
2752
2753 dig100 -= ascii_zero;
2754 dig10 -= ascii_zero;
2755 dig1 -= ascii_zero;
2756 return ((dig100 * 10) + dig10) * 10 + dig1;
2757}
2758
2759/*
2760 * Idem except the pointer to first unparsed byte is returned into <ret> which
2761 * must not be NULL.
2762 */
Willy Tarreau74172752010-10-15 23:21:42 +02002763unsigned int inetaddr_host_lim_ret(char *text, char *stop, char **ret)
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002764{
2765 const unsigned int ascii_zero = ('0' << 24) | ('0' << 16) | ('0' << 8) | '0';
2766 register unsigned int dig100, dig10, dig1;
2767 int s;
Willy Tarreau74172752010-10-15 23:21:42 +02002768 char *p, *d;
Willy Tarreaud54bbdc2009-09-07 11:00:31 +02002769
2770 dig1 = dig10 = dig100 = ascii_zero;
2771 s = 24;
2772
2773 p = text;
2774 while (1) {
2775 if (((unsigned)(*p - '0')) <= 9 && p < stop) {
2776 p++;
2777 continue;
2778 }
2779
2780 /* here, we have a complete byte between <text> and <p> (exclusive) */
2781 if (p == text)
2782 goto end;
2783
2784 d = p - 1;
2785 dig1 |= (unsigned int)(*d << s);
2786 if (d == text)
2787 goto end;
2788
2789 d--;
2790 dig10 |= (unsigned int)(*d << s);
2791 if (d == text)
2792 goto end;
2793
2794 d--;
2795 dig100 |= (unsigned int)(*d << s);
2796 end:
2797 if (!s || p == stop || *p != '.')
2798 break;
2799
2800 s -= 8;
2801 text = ++p;
2802 }
2803
2804 *ret = p;
2805 dig100 -= ascii_zero;
2806 dig10 -= ascii_zero;
2807 dig1 -= ascii_zero;
2808 return ((dig100 * 10) + dig10) * 10 + dig1;
2809}
2810
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002811/* Convert a fixed-length string to an IP address. Returns 0 in case of error,
2812 * or the number of chars read in case of success. Maybe this could be replaced
2813 * by one of the functions above. Also, apparently this function does not support
2814 * hosts above 255 and requires exactly 4 octets.
Willy Tarreau075415a2013-12-12 11:29:39 +01002815 * The destination is only modified on success.
Willy Tarreauf0b38bf2010-06-06 13:22:23 +02002816 */
2817int buf2ip(const char *buf, size_t len, struct in_addr *dst)
2818{
2819 const char *addr;
2820 int saw_digit, octets, ch;
2821 u_char tmp[4], *tp;
2822 const char *cp = buf;
2823
2824 saw_digit = 0;
2825 octets = 0;
2826 *(tp = tmp) = 0;
2827
2828 for (addr = buf; addr - buf < len; addr++) {
2829 unsigned char digit = (ch = *addr) - '0';
2830
2831 if (digit > 9 && ch != '.')
2832 break;
2833
2834 if (digit <= 9) {
2835 u_int new = *tp * 10 + digit;
2836
2837 if (new > 255)
2838 return 0;
2839
2840 *tp = new;
2841
2842 if (!saw_digit) {
2843 if (++octets > 4)
2844 return 0;
2845 saw_digit = 1;
2846 }
2847 } else if (ch == '.' && saw_digit) {
2848 if (octets == 4)
2849 return 0;
2850
2851 *++tp = 0;
2852 saw_digit = 0;
2853 } else
2854 return 0;
2855 }
2856
2857 if (octets < 4)
2858 return 0;
2859
2860 memcpy(&dst->s_addr, tmp, 4);
2861 return addr - cp;
2862}
2863
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002864/* This function converts the string in <buf> of the len <len> to
2865 * struct in6_addr <dst> which must be allocated by the caller.
2866 * This function returns 1 in success case, otherwise zero.
Willy Tarreau075415a2013-12-12 11:29:39 +01002867 * The destination is only modified on success.
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002868 */
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002869int buf2ip6(const char *buf, size_t len, struct in6_addr *dst)
2870{
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002871 char null_term_ip6[INET6_ADDRSTRLEN + 1];
Willy Tarreau075415a2013-12-12 11:29:39 +01002872 struct in6_addr out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002873
Thierry FOURNIERcd659912013-12-11 12:33:54 +01002874 if (len > INET6_ADDRSTRLEN)
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002875 return 0;
2876
2877 memcpy(null_term_ip6, buf, len);
2878 null_term_ip6[len] = '\0';
2879
Willy Tarreau075415a2013-12-12 11:29:39 +01002880 if (!inet_pton(AF_INET6, null_term_ip6, &out))
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002881 return 0;
2882
Willy Tarreau075415a2013-12-12 11:29:39 +01002883 *dst = out;
Thierry FOURNIERd559dd82013-11-22 16:16:59 +01002884 return 1;
2885}
2886
Willy Tarreauacf95772010-06-14 19:09:21 +02002887/* To be used to quote config arg positions. Returns the short string at <ptr>
2888 * surrounded by simple quotes if <ptr> is valid and non-empty, or "end of line"
2889 * if ptr is NULL or empty. The string is locally allocated.
2890 */
2891const char *quote_arg(const char *ptr)
2892{
Christopher Faulet1bc04c72017-10-29 20:14:08 +01002893 static THREAD_LOCAL char val[32];
Willy Tarreauacf95772010-06-14 19:09:21 +02002894 int i;
2895
2896 if (!ptr || !*ptr)
2897 return "end of line";
2898 val[0] = '\'';
Willy Tarreaude2dd6b2013-01-24 02:14:42 +01002899 for (i = 1; i < sizeof(val) - 2 && *ptr; i++)
Willy Tarreauacf95772010-06-14 19:09:21 +02002900 val[i] = *ptr++;
2901 val[i++] = '\'';
2902 val[i] = '\0';
2903 return val;
2904}
2905
Willy Tarreau5b180202010-07-18 10:40:48 +02002906/* returns an operator among STD_OP_* for string <str> or < 0 if unknown */
2907int get_std_op(const char *str)
2908{
2909 int ret = -1;
2910
2911 if (*str == 'e' && str[1] == 'q')
2912 ret = STD_OP_EQ;
2913 else if (*str == 'n' && str[1] == 'e')
2914 ret = STD_OP_NE;
2915 else if (*str == 'l') {
2916 if (str[1] == 'e') ret = STD_OP_LE;
2917 else if (str[1] == 't') ret = STD_OP_LT;
2918 }
2919 else if (*str == 'g') {
2920 if (str[1] == 'e') ret = STD_OP_GE;
2921 else if (str[1] == 't') ret = STD_OP_GT;
2922 }
2923
2924 if (ret == -1 || str[2] != '\0')
2925 return -1;
2926 return ret;
2927}
2928
Willy Tarreau4c14eaa2010-11-24 14:01:45 +01002929/* hash a 32-bit integer to another 32-bit integer */
2930unsigned int full_hash(unsigned int a)
2931{
2932 return __full_hash(a);
2933}
2934
Willy Tarreauf3241112019-02-26 09:56:22 +01002935/* Return the bit position in mask <m> of the nth bit set of rank <r>, between
2936 * 0 and LONGBITS-1 included, starting from the left. For example ranks 0,1,2,3
2937 * for mask 0x55 will be 6, 4, 2 and 0 respectively. This algorithm is based on
2938 * a popcount variant and is described here :
2939 * https://graphics.stanford.edu/~seander/bithacks.html
2940 */
2941unsigned int mask_find_rank_bit(unsigned int r, unsigned long m)
2942{
2943 unsigned long a, b, c, d;
2944 unsigned int s;
2945 unsigned int t;
2946
2947 a = m - ((m >> 1) & ~0UL/3);
2948 b = (a & ~0UL/5) + ((a >> 2) & ~0UL/5);
2949 c = (b + (b >> 4)) & ~0UL/0x11;
2950 d = (c + (c >> 8)) & ~0UL/0x101;
2951
2952 r++; // make r be 1..64
2953
2954 t = 0;
2955 s = LONGBITS;
2956 if (s > 32) {
Willy Tarreau9b6be3b2019-03-18 16:31:18 +01002957 unsigned long d2 = (d >> 16) >> 16;
2958 t = d2 + (d2 >> 16);
Willy Tarreauf3241112019-02-26 09:56:22 +01002959 s -= ((t - r) & 256) >> 3; r -= (t & ((t - r) >> 8));
2960 }
2961
2962 t = (d >> (s - 16)) & 0xff;
2963 s -= ((t - r) & 256) >> 4; r -= (t & ((t - r) >> 8));
2964 t = (c >> (s - 8)) & 0xf;
2965 s -= ((t - r) & 256) >> 5; r -= (t & ((t - r) >> 8));
2966 t = (b >> (s - 4)) & 0x7;
2967 s -= ((t - r) & 256) >> 6; r -= (t & ((t - r) >> 8));
2968 t = (a >> (s - 2)) & 0x3;
2969 s -= ((t - r) & 256) >> 7; r -= (t & ((t - r) >> 8));
2970 t = (m >> (s - 1)) & 0x1;
2971 s -= ((t - r) & 256) >> 8;
2972
2973 return s - 1;
2974}
2975
2976/* Same as mask_find_rank_bit() above but makes use of pre-computed bitmaps
2977 * based on <m>, in <a..d>. These ones must be updated whenever <m> changes
2978 * using mask_prep_rank_map() below.
2979 */
2980unsigned int mask_find_rank_bit_fast(unsigned int r, unsigned long m,
2981 unsigned long a, unsigned long b,
2982 unsigned long c, unsigned long d)
2983{
2984 unsigned int s;
2985 unsigned int t;
2986
2987 r++; // make r be 1..64
2988
2989 t = 0;
2990 s = LONGBITS;
2991 if (s > 32) {
Willy Tarreau9b6be3b2019-03-18 16:31:18 +01002992 unsigned long d2 = (d >> 16) >> 16;
2993 t = d2 + (d2 >> 16);
Willy Tarreauf3241112019-02-26 09:56:22 +01002994 s -= ((t - r) & 256) >> 3; r -= (t & ((t - r) >> 8));
2995 }
2996
2997 t = (d >> (s - 16)) & 0xff;
2998 s -= ((t - r) & 256) >> 4; r -= (t & ((t - r) >> 8));
2999 t = (c >> (s - 8)) & 0xf;
3000 s -= ((t - r) & 256) >> 5; r -= (t & ((t - r) >> 8));
3001 t = (b >> (s - 4)) & 0x7;
3002 s -= ((t - r) & 256) >> 6; r -= (t & ((t - r) >> 8));
3003 t = (a >> (s - 2)) & 0x3;
3004 s -= ((t - r) & 256) >> 7; r -= (t & ((t - r) >> 8));
3005 t = (m >> (s - 1)) & 0x1;
3006 s -= ((t - r) & 256) >> 8;
3007
3008 return s - 1;
3009}
3010
3011/* Prepare the bitmaps used by the fast implementation of the find_rank_bit()
3012 * above.
3013 */
3014void mask_prep_rank_map(unsigned long m,
3015 unsigned long *a, unsigned long *b,
3016 unsigned long *c, unsigned long *d)
3017{
3018 *a = m - ((m >> 1) & ~0UL/3);
3019 *b = (*a & ~0UL/5) + ((*a >> 2) & ~0UL/5);
3020 *c = (*b + (*b >> 4)) & ~0UL/0x11;
3021 *d = (*c + (*c >> 8)) & ~0UL/0x101;
3022}
3023
David du Colombier4f92d322011-03-24 11:09:31 +01003024/* Return non-zero if IPv4 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02003025 * otherwise zero. Note that <addr> may not necessarily be aligned
3026 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01003027 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02003028int in_net_ipv4(const void *addr, const struct in_addr *mask, const struct in_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01003029{
Willy Tarreaueec1d382016-07-13 11:59:39 +02003030 struct in_addr addr_copy;
3031
3032 memcpy(&addr_copy, addr, sizeof(addr_copy));
3033 return((addr_copy.s_addr & mask->s_addr) == (net->s_addr & mask->s_addr));
David du Colombier4f92d322011-03-24 11:09:31 +01003034}
3035
3036/* Return non-zero if IPv6 address is part of the network,
Willy Tarreaueec1d382016-07-13 11:59:39 +02003037 * otherwise zero. Note that <addr> may not necessarily be aligned
3038 * while the two other ones must.
David du Colombier4f92d322011-03-24 11:09:31 +01003039 */
Willy Tarreaueec1d382016-07-13 11:59:39 +02003040int in_net_ipv6(const void *addr, const struct in6_addr *mask, const struct in6_addr *net)
David du Colombier4f92d322011-03-24 11:09:31 +01003041{
3042 int i;
Willy Tarreaueec1d382016-07-13 11:59:39 +02003043 struct in6_addr addr_copy;
David du Colombier4f92d322011-03-24 11:09:31 +01003044
Willy Tarreaueec1d382016-07-13 11:59:39 +02003045 memcpy(&addr_copy, addr, sizeof(addr_copy));
David du Colombier4f92d322011-03-24 11:09:31 +01003046 for (i = 0; i < sizeof(struct in6_addr) / sizeof(int); i++)
Willy Tarreaueec1d382016-07-13 11:59:39 +02003047 if (((((int *)&addr_copy)[i] & ((int *)mask)[i])) !=
David du Colombier4f92d322011-03-24 11:09:31 +01003048 (((int *)net)[i] & ((int *)mask)[i]))
3049 return 0;
3050 return 1;
3051}
3052
3053/* RFC 4291 prefix */
3054const char rfc4291_pfx[] = { 0x00, 0x00, 0x00, 0x00,
3055 0x00, 0x00, 0x00, 0x00,
3056 0x00, 0x00, 0xFF, 0xFF };
3057
Joseph Herlant32b83272018-11-15 11:58:28 -08003058/* Map IPv4 address on IPv6 address, as specified in RFC 3513.
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003059 * Input and output may overlap.
3060 */
David du Colombier4f92d322011-03-24 11:09:31 +01003061void v4tov6(struct in6_addr *sin6_addr, struct in_addr *sin_addr)
3062{
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003063 struct in_addr tmp_addr;
3064
3065 tmp_addr.s_addr = sin_addr->s_addr;
David du Colombier4f92d322011-03-24 11:09:31 +01003066 memcpy(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx));
Thierry FOURNIER4a04dc32013-11-28 16:33:15 +01003067 memcpy(sin6_addr->s6_addr+12, &tmp_addr.s_addr, 4);
David du Colombier4f92d322011-03-24 11:09:31 +01003068}
3069
Joseph Herlant32b83272018-11-15 11:58:28 -08003070/* Map IPv6 address on IPv4 address, as specified in RFC 3513.
David du Colombier4f92d322011-03-24 11:09:31 +01003071 * Return true if conversion is possible and false otherwise.
3072 */
3073int v6tov4(struct in_addr *sin_addr, struct in6_addr *sin6_addr)
3074{
3075 if (memcmp(sin6_addr->s6_addr, rfc4291_pfx, sizeof(rfc4291_pfx)) == 0) {
3076 memcpy(&(sin_addr->s_addr), &(sin6_addr->s6_addr[12]),
3077 sizeof(struct in_addr));
3078 return 1;
3079 }
3080
3081 return 0;
3082}
3083
Baptiste Assmann08b24cf2016-01-23 23:39:12 +01003084/* compare two struct sockaddr_storage and return:
3085 * 0 (true) if the addr is the same in both
3086 * 1 (false) if the addr is not the same in both
3087 * -1 (unable) if one of the addr is not AF_INET*
3088 */
3089int ipcmp(struct sockaddr_storage *ss1, struct sockaddr_storage *ss2)
3090{
3091 if ((ss1->ss_family != AF_INET) && (ss1->ss_family != AF_INET6))
3092 return -1;
3093
3094 if ((ss2->ss_family != AF_INET) && (ss2->ss_family != AF_INET6))
3095 return -1;
3096
3097 if (ss1->ss_family != ss2->ss_family)
3098 return 1;
3099
3100 switch (ss1->ss_family) {
3101 case AF_INET:
3102 return memcmp(&((struct sockaddr_in *)ss1)->sin_addr,
3103 &((struct sockaddr_in *)ss2)->sin_addr,
3104 sizeof(struct in_addr)) != 0;
3105 case AF_INET6:
3106 return memcmp(&((struct sockaddr_in6 *)ss1)->sin6_addr,
3107 &((struct sockaddr_in6 *)ss2)->sin6_addr,
3108 sizeof(struct in6_addr)) != 0;
3109 }
3110
3111 return 1;
3112}
3113
Christopher Faulet9553de72021-02-26 09:12:50 +01003114/* compare a struct sockaddr_storage to a struct net_addr and return :
3115 * 0 (true) if <addr> is matching <net>
3116 * 1 (false) if <addr> is not matching <net>
3117 * -1 (unable) if <addr> or <net> is not AF_INET*
3118 */
3119int ipcmp2net(const struct sockaddr_storage *addr, const struct net_addr *net)
3120{
3121 if ((addr->ss_family != AF_INET) && (addr->ss_family != AF_INET6))
3122 return -1;
3123
3124 if ((net->family != AF_INET) && (net->family != AF_INET6))
3125 return -1;
3126
3127 if (addr->ss_family != net->family)
3128 return 1;
3129
3130 if (addr->ss_family == AF_INET &&
3131 (((struct sockaddr_in *)addr)->sin_addr.s_addr & net->addr.v4.mask.s_addr) == net->addr.v4.ip.s_addr)
3132 return 0;
3133 else {
3134 const struct in6_addr *addr6 = &(((const struct sockaddr_in6*)addr)->sin6_addr);
3135 const struct in6_addr *nip6 = &net->addr.v6.ip;
3136 const struct in6_addr *nmask6 = &net->addr.v6.mask;
3137
3138 if ((read_u32(&addr6->s6_addr[0]) & read_u32(&nmask6->s6_addr[0])) == read_u32(&nip6->s6_addr[0]) &&
3139 (read_u32(&addr6->s6_addr[4]) & read_u32(&nmask6->s6_addr[4])) == read_u32(&nip6->s6_addr[4]) &&
3140 (read_u32(&addr6->s6_addr[8]) & read_u32(&nmask6->s6_addr[8])) == read_u32(&nip6->s6_addr[8]) &&
3141 (read_u32(&addr6->s6_addr[12]) & read_u32(&nmask6->s6_addr[12])) == read_u32(&nip6->s6_addr[12]))
3142 return 0;
3143 }
3144
3145 return 1;
3146}
3147
Baptiste Assmann08396c82016-01-31 00:27:17 +01003148/* copy IP address from <source> into <dest>
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003149 * The caller must allocate and clear <dest> before calling.
3150 * The source must be in either AF_INET or AF_INET6 family, or the destination
3151 * address will be undefined. If the destination address used to hold a port,
3152 * it is preserved, so that this function can be used to switch to another
3153 * address family with no risk. Returns a pointer to the destination.
Baptiste Assmann08396c82016-01-31 00:27:17 +01003154 */
3155struct sockaddr_storage *ipcpy(struct sockaddr_storage *source, struct sockaddr_storage *dest)
3156{
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003157 int prev_port;
3158
3159 prev_port = get_net_port(dest);
3160 memset(dest, 0, sizeof(*dest));
Baptiste Assmann08396c82016-01-31 00:27:17 +01003161 dest->ss_family = source->ss_family;
3162
3163 /* copy new addr and apply it */
3164 switch (source->ss_family) {
3165 case AF_INET:
3166 ((struct sockaddr_in *)dest)->sin_addr.s_addr = ((struct sockaddr_in *)source)->sin_addr.s_addr;
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003167 ((struct sockaddr_in *)dest)->sin_port = prev_port;
Baptiste Assmann08396c82016-01-31 00:27:17 +01003168 break;
3169 case AF_INET6:
3170 memcpy(((struct sockaddr_in6 *)dest)->sin6_addr.s6_addr, ((struct sockaddr_in6 *)source)->sin6_addr.s6_addr, sizeof(struct in6_addr));
Willy Tarreaudc3a9e82016-11-04 18:47:01 +01003171 ((struct sockaddr_in6 *)dest)->sin6_port = prev_port;
Baptiste Assmann08396c82016-01-31 00:27:17 +01003172 break;
3173 }
3174
3175 return dest;
3176}
3177
William Lallemand421f5b52012-02-06 18:15:57 +01003178char *human_time(int t, short hz_div) {
3179 static char rv[sizeof("24855d23h")+1]; // longest of "23h59m" and "59m59s"
3180 char *p = rv;
Willy Tarreau761b3d52014-04-14 14:53:06 +02003181 char *end = rv + sizeof(rv);
William Lallemand421f5b52012-02-06 18:15:57 +01003182 int cnt=2; // print two numbers
3183
3184 if (unlikely(t < 0 || hz_div <= 0)) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003185 snprintf(p, end - p, "?");
William Lallemand421f5b52012-02-06 18:15:57 +01003186 return rv;
3187 }
3188
3189 if (unlikely(hz_div > 1))
3190 t /= hz_div;
3191
3192 if (t >= DAY) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003193 p += snprintf(p, end - p, "%dd", t / DAY);
William Lallemand421f5b52012-02-06 18:15:57 +01003194 cnt--;
3195 }
3196
3197 if (cnt && t % DAY / HOUR) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003198 p += snprintf(p, end - p, "%dh", t % DAY / HOUR);
William Lallemand421f5b52012-02-06 18:15:57 +01003199 cnt--;
3200 }
3201
3202 if (cnt && t % HOUR / MINUTE) {
Willy Tarreau761b3d52014-04-14 14:53:06 +02003203 p += snprintf(p, end - p, "%dm", t % HOUR / MINUTE);
William Lallemand421f5b52012-02-06 18:15:57 +01003204 cnt--;
3205 }
3206
3207 if ((cnt && t % MINUTE) || !t) // also display '0s'
Willy Tarreau761b3d52014-04-14 14:53:06 +02003208 p += snprintf(p, end - p, "%ds", t % MINUTE / SEC);
William Lallemand421f5b52012-02-06 18:15:57 +01003209
3210 return rv;
3211}
3212
3213const char *monthname[12] = {
3214 "Jan", "Feb", "Mar", "Apr", "May", "Jun",
3215 "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"
3216};
3217
3218/* date2str_log: write a date in the format :
3219 * sprintf(str, "%02d/%s/%04d:%02d:%02d:%02d.%03d",
3220 * tm.tm_mday, monthname[tm.tm_mon], tm.tm_year+1900,
3221 * tm.tm_hour, tm.tm_min, tm.tm_sec, (int)date.tv_usec/1000);
3222 *
3223 * without using sprintf. return a pointer to the last char written (\0) or
3224 * NULL if there isn't enough space.
3225 */
Willy Tarreauf16cb412018-09-04 19:08:48 +02003226char *date2str_log(char *dst, const struct tm *tm, const struct timeval *date, size_t size)
William Lallemand421f5b52012-02-06 18:15:57 +01003227{
3228
3229 if (size < 25) /* the size is fixed: 24 chars + \0 */
3230 return NULL;
3231
3232 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003233 if (!dst)
3234 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003235 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003236
William Lallemand421f5b52012-02-06 18:15:57 +01003237 memcpy(dst, monthname[tm->tm_mon], 3); // month
3238 dst += 3;
3239 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003240
William Lallemand421f5b52012-02-06 18:15:57 +01003241 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003242 if (!dst)
3243 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003244 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003245
William Lallemand421f5b52012-02-06 18:15:57 +01003246 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003247 if (!dst)
3248 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003249 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003250
William Lallemand421f5b52012-02-06 18:15:57 +01003251 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003252 if (!dst)
3253 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003254 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003255
William Lallemand421f5b52012-02-06 18:15:57 +01003256 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003257 if (!dst)
3258 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003259 *dst++ = '.';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003260
Willy Tarreau7d9421d2020-02-29 09:08:02 +01003261 dst = utoa_pad((unsigned int)(date->tv_usec/1000)%1000, dst, 4); // milliseconds
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003262 if (!dst)
3263 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003264 *dst = '\0';
3265
3266 return dst;
3267}
3268
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003269/* Base year used to compute leap years */
3270#define TM_YEAR_BASE 1900
3271
3272/* Return the difference in seconds between two times (leap seconds are ignored).
3273 * Retrieved from glibc 2.18 source code.
3274 */
3275static int my_tm_diff(const struct tm *a, const struct tm *b)
3276{
3277 /* Compute intervening leap days correctly even if year is negative.
3278 * Take care to avoid int overflow in leap day calculations,
3279 * but it's OK to assume that A and B are close to each other.
3280 */
3281 int a4 = (a->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (a->tm_year & 3);
3282 int b4 = (b->tm_year >> 2) + (TM_YEAR_BASE >> 2) - ! (b->tm_year & 3);
3283 int a100 = a4 / 25 - (a4 % 25 < 0);
3284 int b100 = b4 / 25 - (b4 % 25 < 0);
3285 int a400 = a100 >> 2;
3286 int b400 = b100 >> 2;
3287 int intervening_leap_days = (a4 - b4) - (a100 - b100) + (a400 - b400);
3288 int years = a->tm_year - b->tm_year;
3289 int days = (365 * years + intervening_leap_days
3290 + (a->tm_yday - b->tm_yday));
3291 return (60 * (60 * (24 * days + (a->tm_hour - b->tm_hour))
3292 + (a->tm_min - b->tm_min))
3293 + (a->tm_sec - b->tm_sec));
3294}
3295
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003296/* Return the GMT offset for a specific local time.
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003297 * Both t and tm must represent the same time.
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003298 * The string returned has the same format as returned by strftime(... "%z", tm).
3299 * Offsets are kept in an internal cache for better performances.
3300 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003301const char *get_gmt_offset(time_t t, struct tm *tm)
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003302{
3303 /* Cache offsets from GMT (depending on whether DST is active or not) */
Christopher Faulet1bc04c72017-10-29 20:14:08 +01003304 static THREAD_LOCAL char gmt_offsets[2][5+1] = { "", "" };
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003305
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003306 char *gmt_offset;
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003307 struct tm tm_gmt;
3308 int diff;
3309 int isdst = tm->tm_isdst;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003310
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003311 /* Pretend DST not active if its status is unknown */
3312 if (isdst < 0)
3313 isdst = 0;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003314
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003315 /* Fetch the offset and initialize it if needed */
3316 gmt_offset = gmt_offsets[isdst & 0x01];
3317 if (unlikely(!*gmt_offset)) {
3318 get_gmtime(t, &tm_gmt);
3319 diff = my_tm_diff(tm, &tm_gmt);
3320 if (diff < 0) {
3321 diff = -diff;
3322 *gmt_offset = '-';
3323 } else {
3324 *gmt_offset = '+';
3325 }
Willy Tarreaue112c8a2019-10-29 10:16:11 +01003326 diff %= 86400U;
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003327 diff /= 60; /* Convert to minutes */
3328 snprintf(gmt_offset+1, 4+1, "%02d%02d", diff/60, diff%60);
3329 }
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003330
Willy Tarreaue112c8a2019-10-29 10:16:11 +01003331 return gmt_offset;
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003332}
3333
William Lallemand421f5b52012-02-06 18:15:57 +01003334/* gmt2str_log: write a date in the format :
3335 * "%02d/%s/%04d:%02d:%02d:%02d +0000" without using snprintf
3336 * return a pointer to the last char written (\0) or
3337 * NULL if there isn't enough space.
3338 */
3339char *gmt2str_log(char *dst, struct tm *tm, size_t size)
3340{
Yuxans Yao4e25b012012-10-19 10:36:09 +08003341 if (size < 27) /* the size is fixed: 26 chars + \0 */
William Lallemand421f5b52012-02-06 18:15:57 +01003342 return NULL;
3343
3344 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003345 if (!dst)
3346 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003347 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003348
William Lallemand421f5b52012-02-06 18:15:57 +01003349 memcpy(dst, monthname[tm->tm_mon], 3); // month
3350 dst += 3;
3351 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003352
William Lallemand421f5b52012-02-06 18:15:57 +01003353 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003354 if (!dst)
3355 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003356 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003357
William Lallemand421f5b52012-02-06 18:15:57 +01003358 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003359 if (!dst)
3360 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003361 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003362
William Lallemand421f5b52012-02-06 18:15:57 +01003363 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003364 if (!dst)
3365 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003366 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003367
William Lallemand421f5b52012-02-06 18:15:57 +01003368 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003369 if (!dst)
3370 return NULL;
William Lallemand421f5b52012-02-06 18:15:57 +01003371 *dst++ = ' ';
3372 *dst++ = '+';
3373 *dst++ = '0';
3374 *dst++ = '0';
3375 *dst++ = '0';
3376 *dst++ = '0';
3377 *dst = '\0';
3378
3379 return dst;
3380}
3381
Yuxans Yao4e25b012012-10-19 10:36:09 +08003382/* localdate2str_log: write a date in the format :
3383 * "%02d/%s/%04d:%02d:%02d:%02d +0000(local timezone)" without using snprintf
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003384 * Both t and tm must represent the same time.
3385 * return a pointer to the last char written (\0) or
3386 * NULL if there isn't enough space.
Yuxans Yao4e25b012012-10-19 10:36:09 +08003387 */
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003388char *localdate2str_log(char *dst, time_t t, struct tm *tm, size_t size)
Yuxans Yao4e25b012012-10-19 10:36:09 +08003389{
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003390 const char *gmt_offset;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003391 if (size < 27) /* the size is fixed: 26 chars + \0 */
3392 return NULL;
3393
Benoit GARNIERe2e5bde2016-03-27 03:04:16 +02003394 gmt_offset = get_gmt_offset(t, tm);
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003395
Yuxans Yao4e25b012012-10-19 10:36:09 +08003396 dst = utoa_pad((unsigned int)tm->tm_mday, dst, 3); // day
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003397 if (!dst)
3398 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003399 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003400
Yuxans Yao4e25b012012-10-19 10:36:09 +08003401 memcpy(dst, monthname[tm->tm_mon], 3); // month
3402 dst += 3;
3403 *dst++ = '/';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003404
Yuxans Yao4e25b012012-10-19 10:36:09 +08003405 dst = utoa_pad((unsigned int)tm->tm_year+1900, dst, 5); // year
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003406 if (!dst)
3407 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003408 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003409
Yuxans Yao4e25b012012-10-19 10:36:09 +08003410 dst = utoa_pad((unsigned int)tm->tm_hour, dst, 3); // hour
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003411 if (!dst)
3412 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003413 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003414
Yuxans Yao4e25b012012-10-19 10:36:09 +08003415 dst = utoa_pad((unsigned int)tm->tm_min, dst, 3); // minutes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003416 if (!dst)
3417 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003418 *dst++ = ':';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003419
Yuxans Yao4e25b012012-10-19 10:36:09 +08003420 dst = utoa_pad((unsigned int)tm->tm_sec, dst, 3); // secondes
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003421 if (!dst)
3422 return NULL;
Yuxans Yao4e25b012012-10-19 10:36:09 +08003423 *dst++ = ' ';
Willy Tarreau4eee38a2019-02-12 11:26:29 +01003424
Benoit GARNIERb413c2a2016-03-27 11:08:03 +02003425 memcpy(dst, gmt_offset, 5); // Offset from local time to GMT
Yuxans Yao4e25b012012-10-19 10:36:09 +08003426 dst += 5;
3427 *dst = '\0';
3428
3429 return dst;
3430}
3431
Willy Tarreaucb1949b2017-07-19 19:05:29 +02003432/* Returns the number of seconds since 01/01/1970 0:0:0 GMT for GMT date <tm>.
3433 * It is meant as a portable replacement for timegm() for use with valid inputs.
3434 * Returns undefined results for invalid dates (eg: months out of range 0..11).
3435 */
3436time_t my_timegm(const struct tm *tm)
3437{
3438 /* Each month has 28, 29, 30 or 31 days, or 28+N. The date in the year
3439 * is thus (current month - 1)*28 + cumulated_N[month] to count the
3440 * sum of the extra N days for elapsed months. The sum of all these N
3441 * days doesn't exceed 30 for a complete year (366-12*28) so it fits
3442 * in a 5-bit word. This means that with 60 bits we can represent a
3443 * matrix of all these values at once, which is fast and efficient to
3444 * access. The extra February day for leap years is not counted here.
3445 *
3446 * Jan : none = 0 (0)
3447 * Feb : Jan = 3 (3)
3448 * Mar : Jan..Feb = 3 (3 + 0)
3449 * Apr : Jan..Mar = 6 (3 + 0 + 3)
3450 * May : Jan..Apr = 8 (3 + 0 + 3 + 2)
3451 * Jun : Jan..May = 11 (3 + 0 + 3 + 2 + 3)
3452 * Jul : Jan..Jun = 13 (3 + 0 + 3 + 2 + 3 + 2)
3453 * Aug : Jan..Jul = 16 (3 + 0 + 3 + 2 + 3 + 2 + 3)
3454 * Sep : Jan..Aug = 19 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3)
3455 * Oct : Jan..Sep = 21 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2)
3456 * Nov : Jan..Oct = 24 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2 + 3)
3457 * Dec : Jan..Nov = 26 (3 + 0 + 3 + 2 + 3 + 2 + 3 + 3 + 2 + 3 + 2)
3458 */
3459 uint64_t extra =
3460 ( 0ULL << 0*5) + ( 3ULL << 1*5) + ( 3ULL << 2*5) + /* Jan, Feb, Mar, */
3461 ( 6ULL << 3*5) + ( 8ULL << 4*5) + (11ULL << 5*5) + /* Apr, May, Jun, */
3462 (13ULL << 6*5) + (16ULL << 7*5) + (19ULL << 8*5) + /* Jul, Aug, Sep, */
3463 (21ULL << 9*5) + (24ULL << 10*5) + (26ULL << 11*5); /* Oct, Nov, Dec, */
3464
3465 unsigned int y = tm->tm_year + 1900;
3466 unsigned int m = tm->tm_mon;
3467 unsigned long days = 0;
3468
3469 /* days since 1/1/1970 for full years */
3470 days += days_since_zero(y) - days_since_zero(1970);
3471
3472 /* days for full months in the current year */
3473 days += 28 * m + ((extra >> (m * 5)) & 0x1f);
3474
3475 /* count + 1 after March for leap years. A leap year is a year multiple
3476 * of 4, unless it's multiple of 100 without being multiple of 400. 2000
3477 * is leap, 1900 isn't, 1904 is.
3478 */
3479 if ((m > 1) && !(y & 3) && ((y % 100) || !(y % 400)))
3480 days++;
3481
3482 days += tm->tm_mday - 1;
3483 return days * 86400ULL + tm->tm_hour * 3600 + tm->tm_min * 60 + tm->tm_sec;
3484}
3485
Thierry Fournier93127942016-01-20 18:49:45 +01003486/* This function check a char. It returns true and updates
3487 * <date> and <len> pointer to the new position if the
3488 * character is found.
3489 */
3490static inline int parse_expect_char(const char **date, int *len, char c)
3491{
3492 if (*len < 1 || **date != c)
3493 return 0;
3494 (*len)--;
3495 (*date)++;
3496 return 1;
3497}
3498
3499/* This function expects a string <str> of len <l>. It return true and updates.
3500 * <date> and <len> if the string matches, otherwise, it returns false.
3501 */
3502static inline int parse_strcmp(const char **date, int *len, char *str, int l)
3503{
3504 if (*len < l || strncmp(*date, str, l) != 0)
3505 return 0;
3506 (*len) -= l;
3507 (*date) += l;
3508 return 1;
3509}
3510
3511/* This macro converts 3 chars name in integer. */
3512#define STR2I3(__a, __b, __c) ((__a) * 65536 + (__b) * 256 + (__c))
3513
3514/* day-name = %x4D.6F.6E ; "Mon", case-sensitive
3515 * / %x54.75.65 ; "Tue", case-sensitive
3516 * / %x57.65.64 ; "Wed", case-sensitive
3517 * / %x54.68.75 ; "Thu", case-sensitive
3518 * / %x46.72.69 ; "Fri", case-sensitive
3519 * / %x53.61.74 ; "Sat", case-sensitive
3520 * / %x53.75.6E ; "Sun", case-sensitive
3521 *
3522 * This array must be alphabetically sorted
3523 */
3524static inline int parse_http_dayname(const char **date, int *len, struct tm *tm)
3525{
3526 if (*len < 3)
3527 return 0;
3528 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3529 case STR2I3('M','o','n'): tm->tm_wday = 1; break;
3530 case STR2I3('T','u','e'): tm->tm_wday = 2; break;
3531 case STR2I3('W','e','d'): tm->tm_wday = 3; break;
3532 case STR2I3('T','h','u'): tm->tm_wday = 4; break;
3533 case STR2I3('F','r','i'): tm->tm_wday = 5; break;
3534 case STR2I3('S','a','t'): tm->tm_wday = 6; break;
3535 case STR2I3('S','u','n'): tm->tm_wday = 7; break;
3536 default: return 0;
3537 }
3538 *len -= 3;
3539 *date += 3;
3540 return 1;
3541}
3542
3543/* month = %x4A.61.6E ; "Jan", case-sensitive
3544 * / %x46.65.62 ; "Feb", case-sensitive
3545 * / %x4D.61.72 ; "Mar", case-sensitive
3546 * / %x41.70.72 ; "Apr", case-sensitive
3547 * / %x4D.61.79 ; "May", case-sensitive
3548 * / %x4A.75.6E ; "Jun", case-sensitive
3549 * / %x4A.75.6C ; "Jul", case-sensitive
3550 * / %x41.75.67 ; "Aug", case-sensitive
3551 * / %x53.65.70 ; "Sep", case-sensitive
3552 * / %x4F.63.74 ; "Oct", case-sensitive
3553 * / %x4E.6F.76 ; "Nov", case-sensitive
3554 * / %x44.65.63 ; "Dec", case-sensitive
3555 *
3556 * This array must be alphabetically sorted
3557 */
3558static inline int parse_http_monthname(const char **date, int *len, struct tm *tm)
3559{
3560 if (*len < 3)
3561 return 0;
3562 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3563 case STR2I3('J','a','n'): tm->tm_mon = 0; break;
3564 case STR2I3('F','e','b'): tm->tm_mon = 1; break;
3565 case STR2I3('M','a','r'): tm->tm_mon = 2; break;
3566 case STR2I3('A','p','r'): tm->tm_mon = 3; break;
3567 case STR2I3('M','a','y'): tm->tm_mon = 4; break;
3568 case STR2I3('J','u','n'): tm->tm_mon = 5; break;
3569 case STR2I3('J','u','l'): tm->tm_mon = 6; break;
3570 case STR2I3('A','u','g'): tm->tm_mon = 7; break;
3571 case STR2I3('S','e','p'): tm->tm_mon = 8; break;
3572 case STR2I3('O','c','t'): tm->tm_mon = 9; break;
3573 case STR2I3('N','o','v'): tm->tm_mon = 10; break;
3574 case STR2I3('D','e','c'): tm->tm_mon = 11; break;
3575 default: return 0;
3576 }
3577 *len -= 3;
3578 *date += 3;
3579 return 1;
3580}
3581
3582/* day-name-l = %x4D.6F.6E.64.61.79 ; "Monday", case-sensitive
3583 * / %x54.75.65.73.64.61.79 ; "Tuesday", case-sensitive
3584 * / %x57.65.64.6E.65.73.64.61.79 ; "Wednesday", case-sensitive
3585 * / %x54.68.75.72.73.64.61.79 ; "Thursday", case-sensitive
3586 * / %x46.72.69.64.61.79 ; "Friday", case-sensitive
3587 * / %x53.61.74.75.72.64.61.79 ; "Saturday", case-sensitive
3588 * / %x53.75.6E.64.61.79 ; "Sunday", case-sensitive
3589 *
3590 * This array must be alphabetically sorted
3591 */
3592static inline int parse_http_ldayname(const char **date, int *len, struct tm *tm)
3593{
3594 if (*len < 6) /* Minimum length. */
3595 return 0;
3596 switch (STR2I3((*date)[0], (*date)[1], (*date)[2])) {
3597 case STR2I3('M','o','n'):
3598 RET0_UNLESS(parse_strcmp(date, len, "Monday", 6));
3599 tm->tm_wday = 1;
3600 return 1;
3601 case STR2I3('T','u','e'):
3602 RET0_UNLESS(parse_strcmp(date, len, "Tuesday", 7));
3603 tm->tm_wday = 2;
3604 return 1;
3605 case STR2I3('W','e','d'):
3606 RET0_UNLESS(parse_strcmp(date, len, "Wednesday", 9));
3607 tm->tm_wday = 3;
3608 return 1;
3609 case STR2I3('T','h','u'):
3610 RET0_UNLESS(parse_strcmp(date, len, "Thursday", 8));
3611 tm->tm_wday = 4;
3612 return 1;
3613 case STR2I3('F','r','i'):
3614 RET0_UNLESS(parse_strcmp(date, len, "Friday", 6));
3615 tm->tm_wday = 5;
3616 return 1;
3617 case STR2I3('S','a','t'):
3618 RET0_UNLESS(parse_strcmp(date, len, "Saturday", 8));
3619 tm->tm_wday = 6;
3620 return 1;
3621 case STR2I3('S','u','n'):
3622 RET0_UNLESS(parse_strcmp(date, len, "Sunday", 6));
3623 tm->tm_wday = 7;
3624 return 1;
3625 }
3626 return 0;
3627}
3628
3629/* This function parses exactly 1 digit and returns the numeric value in "digit". */
3630static inline int parse_digit(const char **date, int *len, int *digit)
3631{
3632 if (*len < 1 || **date < '0' || **date > '9')
3633 return 0;
3634 *digit = (**date - '0');
3635 (*date)++;
3636 (*len)--;
3637 return 1;
3638}
3639
3640/* This function parses exactly 2 digits and returns the numeric value in "digit". */
3641static inline int parse_2digit(const char **date, int *len, int *digit)
3642{
3643 int value;
3644
3645 RET0_UNLESS(parse_digit(date, len, &value));
3646 (*digit) = value * 10;
3647 RET0_UNLESS(parse_digit(date, len, &value));
3648 (*digit) += value;
3649
3650 return 1;
3651}
3652
3653/* This function parses exactly 4 digits and returns the numeric value in "digit". */
3654static inline int parse_4digit(const char **date, int *len, int *digit)
3655{
3656 int value;
3657
3658 RET0_UNLESS(parse_digit(date, len, &value));
3659 (*digit) = value * 1000;
3660
3661 RET0_UNLESS(parse_digit(date, len, &value));
3662 (*digit) += value * 100;
3663
3664 RET0_UNLESS(parse_digit(date, len, &value));
3665 (*digit) += value * 10;
3666
3667 RET0_UNLESS(parse_digit(date, len, &value));
3668 (*digit) += value;
3669
3670 return 1;
3671}
3672
3673/* time-of-day = hour ":" minute ":" second
3674 * ; 00:00:00 - 23:59:60 (leap second)
3675 *
3676 * hour = 2DIGIT
3677 * minute = 2DIGIT
3678 * second = 2DIGIT
3679 */
3680static inline int parse_http_time(const char **date, int *len, struct tm *tm)
3681{
3682 RET0_UNLESS(parse_2digit(date, len, &tm->tm_hour)); /* hour 2DIGIT */
3683 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
3684 RET0_UNLESS(parse_2digit(date, len, &tm->tm_min)); /* min 2DIGIT */
3685 RET0_UNLESS(parse_expect_char(date, len, ':')); /* expect ":" */
3686 RET0_UNLESS(parse_2digit(date, len, &tm->tm_sec)); /* sec 2DIGIT */
3687 return 1;
3688}
3689
3690/* From RFC7231
3691 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3692 *
3693 * IMF-fixdate = day-name "," SP date1 SP time-of-day SP GMT
3694 * ; fixed length/zone/capitalization subset of the format
3695 * ; see Section 3.3 of [RFC5322]
3696 *
3697 *
3698 * date1 = day SP month SP year
3699 * ; e.g., 02 Jun 1982
3700 *
3701 * day = 2DIGIT
3702 * year = 4DIGIT
3703 *
3704 * GMT = %x47.4D.54 ; "GMT", case-sensitive
3705 *
3706 * time-of-day = hour ":" minute ":" second
3707 * ; 00:00:00 - 23:59:60 (leap second)
3708 *
3709 * hour = 2DIGIT
3710 * minute = 2DIGIT
3711 * second = 2DIGIT
3712 *
3713 * DIGIT = decimal 0-9
3714 */
3715int parse_imf_date(const char *date, int len, struct tm *tm)
3716{
David Carlier327298c2016-11-20 10:42:38 +00003717 /* tm_gmtoff, if present, ought to be zero'ed */
3718 memset(tm, 0, sizeof(*tm));
3719
Thierry Fournier93127942016-01-20 18:49:45 +01003720 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
3721 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
3722 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3723 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
3724 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3725 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
3726 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3727 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
3728 tm->tm_year -= 1900;
3729 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3730 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3731 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3732 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
3733 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003734 return 1;
3735}
3736
3737/* From RFC7231
3738 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3739 *
3740 * rfc850-date = day-name-l "," SP date2 SP time-of-day SP GMT
3741 * date2 = day "-" month "-" 2DIGIT
3742 * ; e.g., 02-Jun-82
3743 *
3744 * day = 2DIGIT
3745 */
3746int parse_rfc850_date(const char *date, int len, struct tm *tm)
3747{
3748 int year;
3749
David Carlier327298c2016-11-20 10:42:38 +00003750 /* tm_gmtoff, if present, ought to be zero'ed */
3751 memset(tm, 0, sizeof(*tm));
3752
Thierry Fournier93127942016-01-20 18:49:45 +01003753 RET0_UNLESS(parse_http_ldayname(&date, &len, tm)); /* Read the day name */
3754 RET0_UNLESS(parse_expect_char(&date, &len, ',')); /* expect "," */
3755 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3756 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday)); /* day 2DIGIT */
3757 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
3758 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* Month */
3759 RET0_UNLESS(parse_expect_char(&date, &len, '-')); /* expect "-" */
3760
3761 /* year = 2DIGIT
3762 *
3763 * Recipients of a timestamp value in rfc850-(*date) format, which uses a
3764 * two-digit year, MUST interpret a timestamp that appears to be more
3765 * than 50 years in the future as representing the most recent year in
3766 * the past that had the same last two digits.
3767 */
3768 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_year));
3769
3770 /* expect SP */
3771 if (!parse_expect_char(&date, &len, ' ')) {
3772 /* Maybe we have the date with 4 digits. */
3773 RET0_UNLESS(parse_2digit(&date, &len, &year));
3774 tm->tm_year = (tm->tm_year * 100 + year) - 1900;
3775 /* expect SP */
3776 RET0_UNLESS(parse_expect_char(&date, &len, ' '));
3777 } else {
3778 /* I fix 60 as pivot: >60: +1900, <60: +2000. Note that the
3779 * tm_year is the number of year since 1900, so for +1900, we
3780 * do nothing, and for +2000, we add 100.
3781 */
3782 if (tm->tm_year <= 60)
3783 tm->tm_year += 100;
3784 }
3785
3786 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3787 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3788 RET0_UNLESS(parse_strcmp(&date, &len, "GMT", 3)); /* GMT = %x47.4D.54 ; "GMT", case-sensitive */
3789 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003790
3791 return 1;
3792}
3793
3794/* From RFC7231
3795 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3796 *
3797 * asctime-date = day-name SP date3 SP time-of-day SP year
3798 * date3 = month SP ( 2DIGIT / ( SP 1DIGIT ))
3799 * ; e.g., Jun 2
3800 *
3801 * HTTP-date is case sensitive. A sender MUST NOT generate additional
3802 * whitespace in an HTTP-date beyond that specifically included as SP in
3803 * the grammar.
3804 */
3805int parse_asctime_date(const char *date, int len, struct tm *tm)
3806{
David Carlier327298c2016-11-20 10:42:38 +00003807 /* tm_gmtoff, if present, ought to be zero'ed */
3808 memset(tm, 0, sizeof(*tm));
3809
Thierry Fournier93127942016-01-20 18:49:45 +01003810 RET0_UNLESS(parse_http_dayname(&date, &len, tm)); /* day-name */
3811 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3812 RET0_UNLESS(parse_http_monthname(&date, &len, tm)); /* expect month */
3813 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3814
3815 /* expect SP and 1DIGIT or 2DIGIT */
3816 if (parse_expect_char(&date, &len, ' '))
3817 RET0_UNLESS(parse_digit(&date, &len, &tm->tm_mday));
3818 else
3819 RET0_UNLESS(parse_2digit(&date, &len, &tm->tm_mday));
3820
3821 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3822 RET0_UNLESS(parse_http_time(&date, &len, tm)); /* Parse time. */
3823 RET0_UNLESS(parse_expect_char(&date, &len, ' ')); /* expect SP */
3824 RET0_UNLESS(parse_4digit(&date, &len, &tm->tm_year)); /* year = 4DIGIT */
3825 tm->tm_year -= 1900;
3826 tm->tm_isdst = -1;
Thierry Fournier93127942016-01-20 18:49:45 +01003827 return 1;
3828}
3829
3830/* From RFC7231
3831 * https://tools.ietf.org/html/rfc7231#section-7.1.1.1
3832 *
3833 * HTTP-date = IMF-fixdate / obs-date
3834 * obs-date = rfc850-date / asctime-date
3835 *
3836 * parses an HTTP date in the RFC format and is accepted
3837 * alternatives. <date> is the strinf containing the date,
3838 * len is the len of the string. <tm> is filled with the
3839 * parsed time. We must considers this time as GMT.
3840 */
3841int parse_http_date(const char *date, int len, struct tm *tm)
3842{
3843 if (parse_imf_date(date, len, tm))
3844 return 1;
3845
3846 if (parse_rfc850_date(date, len, tm))
3847 return 1;
3848
3849 if (parse_asctime_date(date, len, tm))
3850 return 1;
3851
3852 return 0;
3853}
3854
Willy Tarreau4deeb102021-01-29 10:47:52 +01003855/* print the time <ns> in a short form (exactly 7 chars) at the end of buffer
3856 * <out>. "-" is printed if the value is zero, "inf" if larger than 1000 years.
3857 * It returns the new buffer length, or 0 if it doesn't fit. The value will be
3858 * surrounded by <pfx> and <sfx> respectively if not NULL.
3859 */
3860int print_time_short(struct buffer *out, const char *pfx, uint64_t ns, const char *sfx)
3861{
3862 double val = ns; // 52 bits of mantissa keep ns accuracy over 52 days
3863 const char *unit;
3864
3865 if (!pfx)
3866 pfx = "";
3867 if (!sfx)
3868 sfx = "";
3869
3870 do {
3871 unit = " - "; if (val <= 0.0) break;
3872 unit = "ns"; if (val < 1000.0) break;
3873 unit = "us"; val /= 1000.0; if (val < 1000.0) break;
3874 unit = "ms"; val /= 1000.0; if (val < 1000.0) break;
3875 unit = "s "; val /= 1000.0; if (val < 60.0) break;
3876 unit = "m "; val /= 60.0; if (val < 60.0) break;
3877 unit = "h "; val /= 60.0; if (val < 24.0) break;
3878 unit = "d "; val /= 24.0; if (val < 365.0) break;
3879 unit = "yr"; val /= 365.0; if (val < 1000.0) break;
3880 unit = " inf "; val = 0.0; break;
3881 } while (0);
3882
3883 if (val <= 0.0)
3884 return chunk_appendf(out, "%s%7s%s", pfx, unit, sfx);
3885 else if (val < 10.0)
3886 return chunk_appendf(out, "%s%1.3f%s%s", pfx, val, unit, sfx);
3887 else if (val < 100.0)
3888 return chunk_appendf(out, "%s%2.2f%s%s", pfx, val, unit, sfx);
3889 else
3890 return chunk_appendf(out, "%s%3.1f%s%s", pfx, val, unit, sfx);
3891}
3892
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003893/* Dynamically allocates a string of the proper length to hold the formatted
3894 * output. NULL is returned on error. The caller is responsible for freeing the
3895 * memory area using free(). The resulting string is returned in <out> if the
3896 * pointer is not NULL. A previous version of <out> might be used to build the
3897 * new string, and it will be freed before returning if it is not NULL, which
3898 * makes it possible to build complex strings from iterative calls without
3899 * having to care about freeing intermediate values, as in the example below :
3900 *
3901 * memprintf(&err, "invalid argument: '%s'", arg);
3902 * ...
3903 * memprintf(&err, "parser said : <%s>\n", *err);
3904 * ...
3905 * free(*err);
3906 *
3907 * This means that <err> must be initialized to NULL before first invocation.
3908 * The return value also holds the allocated string, which eases error checking
3909 * and immediate consumption. If the output pointer is not used, NULL must be
Willy Tarreaueb6cead2012-09-20 19:43:14 +02003910 * passed instead and it will be ignored. The returned message will then also
3911 * be NULL so that the caller does not have to bother with freeing anything.
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003912 *
3913 * It is also convenient to use it without any free except the last one :
3914 * err = NULL;
3915 * if (!fct1(err)) report(*err);
3916 * if (!fct2(err)) report(*err);
3917 * if (!fct3(err)) report(*err);
3918 * free(*err);
Christopher Faulet93a518f2017-10-24 11:25:33 +02003919 *
3920 * memprintf relies on memvprintf. This last version can be called from any
3921 * function with variadic arguments.
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003922 */
Christopher Faulet93a518f2017-10-24 11:25:33 +02003923char *memvprintf(char **out, const char *format, va_list orig_args)
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003924{
3925 va_list args;
3926 char *ret = NULL;
3927 int allocated = 0;
3928 int needed = 0;
3929
Willy Tarreaueb6cead2012-09-20 19:43:14 +02003930 if (!out)
3931 return NULL;
3932
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003933 do {
Willy Tarreaue0609f52019-03-29 19:13:23 +01003934 char buf1;
3935
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003936 /* vsnprintf() will return the required length even when the
3937 * target buffer is NULL. We do this in a loop just in case
3938 * intermediate evaluations get wrong.
3939 */
Christopher Faulet93a518f2017-10-24 11:25:33 +02003940 va_copy(args, orig_args);
Willy Tarreaue0609f52019-03-29 19:13:23 +01003941 needed = vsnprintf(ret ? ret : &buf1, allocated, format, args);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003942 va_end(args);
Willy Tarreau1b2fed62013-04-01 22:48:54 +02003943 if (needed < allocated) {
3944 /* Note: on Solaris 8, the first iteration always
3945 * returns -1 if allocated is zero, so we force a
3946 * retry.
3947 */
3948 if (!allocated)
3949 needed = 0;
3950 else
3951 break;
3952 }
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003953
Willy Tarreau1b2fed62013-04-01 22:48:54 +02003954 allocated = needed + 1;
Hubert Verstraete831962e2016-06-28 22:44:26 +02003955 ret = my_realloc2(ret, allocated);
Willy Tarreau9a7bea52012-04-27 11:16:50 +02003956 } while (ret);
3957
3958 if (needed < 0) {
3959 /* an error was encountered */
3960 free(ret);
3961 ret = NULL;
3962 }
3963
3964 if (out) {
3965 free(*out);
3966 *out = ret;
3967 }
3968
3969 return ret;
3970}
William Lallemand421f5b52012-02-06 18:15:57 +01003971
Christopher Faulet93a518f2017-10-24 11:25:33 +02003972char *memprintf(char **out, const char *format, ...)
3973{
3974 va_list args;
3975 char *ret = NULL;
3976
3977 va_start(args, format);
3978 ret = memvprintf(out, format, args);
3979 va_end(args);
3980
3981 return ret;
3982}
3983
Willy Tarreau21c705b2012-09-14 11:40:36 +02003984/* Used to add <level> spaces before each line of <out>, unless there is only one line.
3985 * The input argument is automatically freed and reassigned. The result will have to be
Willy Tarreau70eec382012-10-10 08:56:47 +02003986 * freed by the caller. It also supports being passed a NULL which results in the same
3987 * output.
Willy Tarreau21c705b2012-09-14 11:40:36 +02003988 * Example of use :
3989 * parse(cmd, &err); (callee: memprintf(&err, ...))
3990 * fprintf(stderr, "Parser said: %s\n", indent_error(&err));
3991 * free(err);
3992 */
3993char *indent_msg(char **out, int level)
3994{
3995 char *ret, *in, *p;
3996 int needed = 0;
3997 int lf = 0;
3998 int lastlf = 0;
3999 int len;
4000
Willy Tarreau70eec382012-10-10 08:56:47 +02004001 if (!out || !*out)
4002 return NULL;
4003
Willy Tarreau21c705b2012-09-14 11:40:36 +02004004 in = *out - 1;
4005 while ((in = strchr(in + 1, '\n')) != NULL) {
4006 lastlf = in - *out;
4007 lf++;
4008 }
4009
4010 if (!lf) /* single line, no LF, return it as-is */
4011 return *out;
4012
4013 len = strlen(*out);
4014
4015 if (lf == 1 && lastlf == len - 1) {
4016 /* single line, LF at end, strip it and return as-is */
4017 (*out)[lastlf] = 0;
4018 return *out;
4019 }
4020
4021 /* OK now we have at least one LF, we need to process the whole string
4022 * as a multi-line string. What we'll do :
4023 * - prefix with an LF if there is none
4024 * - add <level> spaces before each line
4025 * This means at most ( 1 + level + (len-lf) + lf*<1+level) ) =
4026 * 1 + level + len + lf * level = 1 + level * (lf + 1) + len.
4027 */
4028
4029 needed = 1 + level * (lf + 1) + len + 1;
4030 p = ret = malloc(needed);
4031 in = *out;
4032
4033 /* skip initial LFs */
4034 while (*in == '\n')
4035 in++;
4036
4037 /* copy each line, prefixed with LF and <level> spaces, and without the trailing LF */
4038 while (*in) {
4039 *p++ = '\n';
4040 memset(p, ' ', level);
4041 p += level;
4042 do {
4043 *p++ = *in++;
4044 } while (*in && *in != '\n');
4045 if (*in)
4046 in++;
4047 }
4048 *p = 0;
4049
4050 free(*out);
4051 *out = ret;
4052
4053 return ret;
4054}
4055
Willy Tarreaua2c99112019-08-21 13:17:37 +02004056/* makes a copy of message <in> into <out>, with each line prefixed with <pfx>
4057 * and end of lines replaced with <eol> if not 0. The first line to indent has
4058 * to be indicated in <first> (starts at zero), so that it is possible to skip
4059 * indenting the first line if it has to be appended after an existing message.
4060 * Empty strings are never indented, and NULL strings are considered empty both
4061 * for <in> and <pfx>. It returns non-zero if an EOL was appended as the last
4062 * character, non-zero otherwise.
4063 */
4064int append_prefixed_str(struct buffer *out, const char *in, const char *pfx, char eol, int first)
4065{
4066 int bol, lf;
4067 int pfxlen = pfx ? strlen(pfx) : 0;
4068
4069 if (!in)
4070 return 0;
4071
4072 bol = 1;
4073 lf = 0;
4074 while (*in) {
4075 if (bol && pfxlen) {
4076 if (first > 0)
4077 first--;
4078 else
4079 b_putblk(out, pfx, pfxlen);
4080 bol = 0;
4081 }
4082
4083 lf = (*in == '\n');
4084 bol |= lf;
4085 b_putchr(out, (lf && eol) ? eol : *in);
4086 in++;
4087 }
4088 return lf;
4089}
4090
Willy Tarreau9d22e562019-03-29 18:49:09 +01004091/* removes environment variable <name> from the environment as found in
4092 * environ. This is only provided as an alternative for systems without
4093 * unsetenv() (old Solaris and AIX versions). THIS IS NOT THREAD SAFE.
Ilya Shipitsin856aabc2020-04-16 23:51:34 +05004094 * The principle is to scan environ for each occurrence of variable name
Willy Tarreau9d22e562019-03-29 18:49:09 +01004095 * <name> and to replace the matching pointers with the last pointer of
4096 * the array (since variables are not ordered).
4097 * It always returns 0 (success).
4098 */
4099int my_unsetenv(const char *name)
4100{
4101 extern char **environ;
4102 char **p = environ;
4103 int vars;
4104 int next;
4105 int len;
4106
4107 len = strlen(name);
4108 for (vars = 0; p[vars]; vars++)
4109 ;
4110 next = 0;
4111 while (next < vars) {
4112 if (strncmp(p[next], name, len) != 0 || p[next][len] != '=') {
4113 next++;
4114 continue;
4115 }
4116 if (next < vars - 1)
4117 p[next] = p[vars - 1];
4118 p[--vars] = NULL;
4119 }
4120 return 0;
4121}
4122
Willy Tarreaudad36a32013-03-11 01:20:04 +01004123/* Convert occurrences of environment variables in the input string to their
4124 * corresponding value. A variable is identified as a series of alphanumeric
4125 * characters or underscores following a '$' sign. The <in> string must be
4126 * free()able. NULL returns NULL. The resulting string might be reallocated if
4127 * some expansion is made. Variable names may also be enclosed into braces if
4128 * needed (eg: to concatenate alphanum characters).
4129 */
4130char *env_expand(char *in)
4131{
4132 char *txt_beg;
4133 char *out;
4134 char *txt_end;
4135 char *var_beg;
4136 char *var_end;
4137 char *value;
4138 char *next;
4139 int out_len;
4140 int val_len;
4141
4142 if (!in)
4143 return in;
4144
4145 value = out = NULL;
4146 out_len = 0;
4147
4148 txt_beg = in;
4149 do {
4150 /* look for next '$' sign in <in> */
4151 for (txt_end = txt_beg; *txt_end && *txt_end != '$'; txt_end++);
4152
4153 if (!*txt_end && !out) /* end and no expansion performed */
4154 return in;
4155
4156 val_len = 0;
4157 next = txt_end;
4158 if (*txt_end == '$') {
4159 char save;
4160
4161 var_beg = txt_end + 1;
4162 if (*var_beg == '{')
4163 var_beg++;
4164
4165 var_end = var_beg;
Willy Tarreau90807112020-02-25 08:16:33 +01004166 while (isalnum((unsigned char)*var_end) || *var_end == '_') {
Willy Tarreaudad36a32013-03-11 01:20:04 +01004167 var_end++;
4168 }
4169
4170 next = var_end;
4171 if (*var_end == '}' && (var_beg > txt_end + 1))
4172 next++;
4173
4174 /* get value of the variable name at this location */
4175 save = *var_end;
4176 *var_end = '\0';
4177 value = getenv(var_beg);
4178 *var_end = save;
4179 val_len = value ? strlen(value) : 0;
4180 }
4181
Hubert Verstraete831962e2016-06-28 22:44:26 +02004182 out = my_realloc2(out, out_len + (txt_end - txt_beg) + val_len + 1);
Willy Tarreaudad36a32013-03-11 01:20:04 +01004183 if (txt_end > txt_beg) {
4184 memcpy(out + out_len, txt_beg, txt_end - txt_beg);
4185 out_len += txt_end - txt_beg;
4186 }
4187 if (val_len) {
4188 memcpy(out + out_len, value, val_len);
4189 out_len += val_len;
4190 }
4191 out[out_len] = 0;
4192 txt_beg = next;
4193 } while (*txt_beg);
4194
4195 /* here we know that <out> was allocated and that we don't need <in> anymore */
4196 free(in);
4197 return out;
4198}
4199
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004200
4201/* same as strstr() but case-insensitive and with limit length */
4202const char *strnistr(const char *str1, int len_str1, const char *str2, int len_str2)
4203{
4204 char *pptr, *sptr, *start;
Willy Tarreauc8746532014-05-28 23:05:07 +02004205 unsigned int slen, plen;
4206 unsigned int tmp1, tmp2;
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004207
4208 if (str1 == NULL || len_str1 == 0) // search pattern into an empty string => search is not found
4209 return NULL;
4210
4211 if (str2 == NULL || len_str2 == 0) // pattern is empty => every str1 match
4212 return str1;
4213
4214 if (len_str1 < len_str2) // pattern is longer than string => search is not found
4215 return NULL;
4216
4217 for (tmp1 = 0, start = (char *)str1, pptr = (char *)str2, slen = len_str1, plen = len_str2; slen >= plen; start++, slen--) {
Willy Tarreauf278eec2020-07-05 21:46:32 +02004218 while (toupper((unsigned char)*start) != toupper((unsigned char)*str2)) {
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004219 start++;
4220 slen--;
4221 tmp1++;
4222
4223 if (tmp1 >= len_str1)
4224 return NULL;
4225
4226 /* if pattern longer than string */
4227 if (slen < plen)
4228 return NULL;
4229 }
4230
4231 sptr = start;
4232 pptr = (char *)str2;
4233
4234 tmp2 = 0;
Willy Tarreauf278eec2020-07-05 21:46:32 +02004235 while (toupper((unsigned char)*sptr) == toupper((unsigned char)*pptr)) {
de Lafond Guillaume88c278f2013-04-15 19:27:10 +02004236 sptr++;
4237 pptr++;
4238 tmp2++;
4239
4240 if (*pptr == '\0' || tmp2 == len_str2) /* end of pattern found */
4241 return start;
4242 if (*sptr == '\0' || tmp2 == len_str1) /* end of string found and the pattern is not fully found */
4243 return NULL;
4244 }
4245 }
4246 return NULL;
4247}
4248
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02004249/* This function read the next valid utf8 char.
4250 * <s> is the byte srray to be decode, <len> is its length.
4251 * The function returns decoded char encoded like this:
4252 * The 4 msb are the return code (UTF8_CODE_*), the 4 lsb
4253 * are the length read. The decoded character is stored in <c>.
4254 */
4255unsigned char utf8_next(const char *s, int len, unsigned int *c)
4256{
4257 const unsigned char *p = (unsigned char *)s;
4258 int dec;
4259 unsigned char code = UTF8_CODE_OK;
4260
4261 if (len < 1)
4262 return UTF8_CODE_OK;
4263
4264 /* Check the type of UTF8 sequence
4265 *
4266 * 0... .... 0x00 <= x <= 0x7f : 1 byte: ascii char
4267 * 10.. .... 0x80 <= x <= 0xbf : invalid sequence
4268 * 110. .... 0xc0 <= x <= 0xdf : 2 bytes
4269 * 1110 .... 0xe0 <= x <= 0xef : 3 bytes
4270 * 1111 0... 0xf0 <= x <= 0xf7 : 4 bytes
4271 * 1111 10.. 0xf8 <= x <= 0xfb : 5 bytes
4272 * 1111 110. 0xfc <= x <= 0xfd : 6 bytes
4273 * 1111 111. 0xfe <= x <= 0xff : invalid sequence
4274 */
4275 switch (*p) {
4276 case 0x00 ... 0x7f:
4277 *c = *p;
4278 return UTF8_CODE_OK | 1;
4279
4280 case 0x80 ... 0xbf:
4281 *c = *p;
4282 return UTF8_CODE_BADSEQ | 1;
4283
4284 case 0xc0 ... 0xdf:
4285 if (len < 2) {
4286 *c = *p;
4287 return UTF8_CODE_BADSEQ | 1;
4288 }
4289 *c = *p & 0x1f;
4290 dec = 1;
4291 break;
4292
4293 case 0xe0 ... 0xef:
4294 if (len < 3) {
4295 *c = *p;
4296 return UTF8_CODE_BADSEQ | 1;
4297 }
4298 *c = *p & 0x0f;
4299 dec = 2;
4300 break;
4301
4302 case 0xf0 ... 0xf7:
4303 if (len < 4) {
4304 *c = *p;
4305 return UTF8_CODE_BADSEQ | 1;
4306 }
4307 *c = *p & 0x07;
4308 dec = 3;
4309 break;
4310
4311 case 0xf8 ... 0xfb:
4312 if (len < 5) {
4313 *c = *p;
4314 return UTF8_CODE_BADSEQ | 1;
4315 }
4316 *c = *p & 0x03;
4317 dec = 4;
4318 break;
4319
4320 case 0xfc ... 0xfd:
4321 if (len < 6) {
4322 *c = *p;
4323 return UTF8_CODE_BADSEQ | 1;
4324 }
4325 *c = *p & 0x01;
4326 dec = 5;
4327 break;
4328
4329 case 0xfe ... 0xff:
4330 default:
4331 *c = *p;
4332 return UTF8_CODE_BADSEQ | 1;
4333 }
4334
4335 p++;
4336
4337 while (dec > 0) {
4338
4339 /* need 0x10 for the 2 first bits */
4340 if ( ( *p & 0xc0 ) != 0x80 )
4341 return UTF8_CODE_BADSEQ | ((p-(unsigned char *)s)&0xffff);
4342
4343 /* add data at char */
4344 *c = ( *c << 6 ) | ( *p & 0x3f );
4345
4346 dec--;
4347 p++;
4348 }
4349
4350 /* Check ovelong encoding.
4351 * 1 byte : 5 + 6 : 11 : 0x80 ... 0x7ff
4352 * 2 bytes : 4 + 6 + 6 : 16 : 0x800 ... 0xffff
4353 * 3 bytes : 3 + 6 + 6 + 6 : 21 : 0x10000 ... 0x1fffff
4354 */
Thierry FOURNIER9e7ec082015-03-12 19:32:38 +01004355 if (( *c <= 0x7f && (p-(unsigned char *)s) > 1) ||
Thierry FOURNIER317e1c42014-08-12 10:20:47 +02004356 (*c >= 0x80 && *c <= 0x7ff && (p-(unsigned char *)s) > 2) ||
4357 (*c >= 0x800 && *c <= 0xffff && (p-(unsigned char *)s) > 3) ||
4358 (*c >= 0x10000 && *c <= 0x1fffff && (p-(unsigned char *)s) > 4))
4359 code |= UTF8_CODE_OVERLONG;
4360
4361 /* Check invalid UTF8 range. */
4362 if ((*c >= 0xd800 && *c <= 0xdfff) ||
4363 (*c >= 0xfffe && *c <= 0xffff))
4364 code |= UTF8_CODE_INVRANGE;
4365
4366 return code | ((p-(unsigned char *)s)&0x0f);
4367}
4368
Maxime de Roucydc887852016-05-13 23:52:54 +02004369/* append a copy of string <str> (in a wordlist) at the end of the list <li>
4370 * On failure : return 0 and <err> filled with an error message.
4371 * The caller is responsible for freeing the <err> and <str> copy
4372 * memory area using free()
4373 */
4374int list_append_word(struct list *li, const char *str, char **err)
4375{
4376 struct wordlist *wl;
4377
4378 wl = calloc(1, sizeof(*wl));
4379 if (!wl) {
4380 memprintf(err, "out of memory");
4381 goto fail_wl;
4382 }
4383
4384 wl->s = strdup(str);
4385 if (!wl->s) {
4386 memprintf(err, "out of memory");
4387 goto fail_wl_s;
4388 }
4389
4390 LIST_ADDQ(li, &wl->list);
4391
4392 return 1;
4393
4394fail_wl_s:
4395 free(wl->s);
4396fail_wl:
4397 free(wl);
4398 return 0;
4399}
4400
Willy Tarreau37101052019-05-20 16:48:20 +02004401/* indicates if a memory location may safely be read or not. The trick consists
4402 * in performing a harmless syscall using this location as an input and letting
4403 * the operating system report whether it's OK or not. For this we have the
4404 * stat() syscall, which will return EFAULT when the memory location supposed
4405 * to contain the file name is not readable. If it is readable it will then
4406 * either return 0 if the area contains an existing file name, or -1 with
4407 * another code. This must not be abused, and some audit systems might detect
4408 * this as abnormal activity. It's used only for unsafe dumps.
4409 */
4410int may_access(const void *ptr)
4411{
4412 struct stat buf;
4413
4414 if (stat(ptr, &buf) == 0)
4415 return 1;
4416 if (errno == EFAULT)
4417 return 0;
4418 return 1;
4419}
4420
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004421/* print a string of text buffer to <out>. The format is :
4422 * Non-printable chars \t, \n, \r and \e are * encoded in C format.
4423 * Other non-printable chars are encoded "\xHH". Space, '\', and '=' are also escaped.
4424 * Print stopped if null char or <bsize> is reached, or if no more place in the chunk.
4425 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004426int dump_text(struct buffer *out, const char *buf, int bsize)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004427{
4428 unsigned char c;
4429 int ptr = 0;
4430
4431 while (buf[ptr] && ptr < bsize) {
4432 c = buf[ptr];
Willy Tarreau90807112020-02-25 08:16:33 +01004433 if (isprint((unsigned char)c) && isascii((unsigned char)c) && c != '\\' && c != ' ' && c != '=') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004434 if (out->data > out->size - 1)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004435 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004436 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004437 }
4438 else if (c == '\t' || c == '\n' || c == '\r' || c == '\e' || c == '\\' || c == ' ' || c == '=') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004439 if (out->data > out->size - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004440 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004441 out->area[out->data++] = '\\';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004442 switch (c) {
4443 case ' ': c = ' '; break;
4444 case '\t': c = 't'; break;
4445 case '\n': c = 'n'; break;
4446 case '\r': c = 'r'; break;
4447 case '\e': c = 'e'; break;
4448 case '\\': c = '\\'; break;
4449 case '=': c = '='; break;
4450 }
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004451 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004452 }
4453 else {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004454 if (out->data > out->size - 4)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004455 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004456 out->area[out->data++] = '\\';
4457 out->area[out->data++] = 'x';
4458 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4459 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004460 }
4461 ptr++;
4462 }
4463
4464 return ptr;
4465}
4466
4467/* print a buffer in hexa.
4468 * Print stopped if <bsize> is reached, or if no more place in the chunk.
4469 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004470int dump_binary(struct buffer *out, const char *buf, int bsize)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004471{
4472 unsigned char c;
4473 int ptr = 0;
4474
4475 while (ptr < bsize) {
4476 c = buf[ptr];
4477
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004478 if (out->data > out->size - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004479 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004480 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4481 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004482
4483 ptr++;
4484 }
4485 return ptr;
4486}
4487
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004488/* Appends into buffer <out> a hex dump of memory area <buf> for <len> bytes,
4489 * prepending each line with prefix <pfx>. The output is *not* initialized.
4490 * The output will not wrap pas the buffer's end so it is more optimal if the
4491 * caller makes sure the buffer is aligned first. A trailing zero will always
4492 * be appended (and not counted) if there is room for it. The caller must make
Willy Tarreau37101052019-05-20 16:48:20 +02004493 * sure that the area is dumpable first. If <unsafe> is non-null, the memory
4494 * locations are checked first for being readable.
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004495 */
Willy Tarreau37101052019-05-20 16:48:20 +02004496void dump_hex(struct buffer *out, const char *pfx, const void *buf, int len, int unsafe)
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004497{
4498 const unsigned char *d = buf;
4499 int i, j, start;
4500
4501 d = (const unsigned char *)(((unsigned long)buf) & -16);
4502 start = ((unsigned long)buf) & 15;
4503
4504 for (i = 0; i < start + len; i += 16) {
4505 chunk_appendf(out, (sizeof(void *) == 4) ? "%s%8p: " : "%s%16p: ", pfx, d + i);
4506
Willy Tarreau37101052019-05-20 16:48:20 +02004507 // 0: unchecked, 1: checked safe, 2: danger
4508 unsafe = !!unsafe;
4509 if (unsafe && !may_access(d + i))
4510 unsafe = 2;
4511
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004512 for (j = 0; j < 16; j++) {
Willy Tarreau37101052019-05-20 16:48:20 +02004513 if ((i + j < start) || (i + j >= start + len))
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004514 chunk_strcat(out, "'' ");
Willy Tarreau37101052019-05-20 16:48:20 +02004515 else if (unsafe > 1)
4516 chunk_strcat(out, "** ");
4517 else
4518 chunk_appendf(out, "%02x ", d[i + j]);
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004519
4520 if (j == 7)
4521 chunk_strcat(out, "- ");
4522 }
4523 chunk_strcat(out, " ");
4524 for (j = 0; j < 16; j++) {
Willy Tarreau37101052019-05-20 16:48:20 +02004525 if ((i + j < start) || (i + j >= start + len))
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004526 chunk_strcat(out, "'");
Willy Tarreau37101052019-05-20 16:48:20 +02004527 else if (unsafe > 1)
4528 chunk_strcat(out, "*");
Willy Tarreau90807112020-02-25 08:16:33 +01004529 else if (isprint((unsigned char)d[i + j]))
Willy Tarreau37101052019-05-20 16:48:20 +02004530 chunk_appendf(out, "%c", d[i + j]);
4531 else
4532 chunk_strcat(out, ".");
Willy Tarreau9fc5dcb2019-05-20 16:13:40 +02004533 }
4534 chunk_strcat(out, "\n");
4535 }
4536}
4537
Willy Tarreau762fb3e2020-03-03 15:57:10 +01004538/* dumps <pfx> followed by <n> bytes from <addr> in hex form into buffer <buf>
4539 * enclosed in brackets after the address itself, formatted on 14 chars
4540 * including the "0x" prefix. This is meant to be used as a prefix for code
4541 * areas. For example:
4542 * "0x7f10b6557690 [48 c7 c0 0f 00 00 00 0f]"
4543 * It relies on may_access() to know if the bytes are dumpable, otherwise "--"
4544 * is emitted. A NULL <pfx> will be considered empty.
4545 */
4546void dump_addr_and_bytes(struct buffer *buf, const char *pfx, const void *addr, int n)
4547{
4548 int ok = 0;
4549 int i;
4550
4551 chunk_appendf(buf, "%s%#14lx [", pfx ? pfx : "", (long)addr);
4552
4553 for (i = 0; i < n; i++) {
4554 if (i == 0 || (((long)(addr + i) ^ (long)(addr)) & 4096))
4555 ok = may_access(addr + i);
4556 if (ok)
4557 chunk_appendf(buf, "%02x%s", ((uint8_t*)addr)[i], (i<n-1) ? " " : "]");
4558 else
4559 chunk_appendf(buf, "--%s", (i<n-1) ? " " : "]");
4560 }
4561}
4562
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004563/* print a line of text buffer (limited to 70 bytes) to <out>. The format is :
4564 * <2 spaces> <offset=5 digits> <space or plus> <space> <70 chars max> <\n>
4565 * which is 60 chars per line. Non-printable chars \t, \n, \r and \e are
4566 * encoded in C format. Other non-printable chars are encoded "\xHH". Original
4567 * lines are respected within the limit of 70 output chars. Lines that are
4568 * continuation of a previous truncated line begin with "+" instead of " "
4569 * after the offset. The new pointer is returned.
4570 */
Willy Tarreau83061a82018-07-13 11:56:34 +02004571int dump_text_line(struct buffer *out, const char *buf, int bsize, int len,
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004572 int *line, int ptr)
4573{
4574 int end;
4575 unsigned char c;
4576
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004577 end = out->data + 80;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004578 if (end > out->size)
4579 return ptr;
4580
4581 chunk_appendf(out, " %05d%c ", ptr, (ptr == *line) ? ' ' : '+');
4582
4583 while (ptr < len && ptr < bsize) {
4584 c = buf[ptr];
Willy Tarreau90807112020-02-25 08:16:33 +01004585 if (isprint((unsigned char)c) && isascii((unsigned char)c) && c != '\\') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004586 if (out->data > end - 2)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004587 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004588 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004589 } else if (c == '\t' || c == '\n' || c == '\r' || c == '\e' || c == '\\') {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004590 if (out->data > end - 3)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004591 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004592 out->area[out->data++] = '\\';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004593 switch (c) {
4594 case '\t': c = 't'; break;
4595 case '\n': c = 'n'; break;
4596 case '\r': c = 'r'; break;
4597 case '\e': c = 'e'; break;
4598 case '\\': c = '\\'; break;
4599 }
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004600 out->area[out->data++] = c;
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004601 } else {
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004602 if (out->data > end - 5)
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004603 break;
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004604 out->area[out->data++] = '\\';
4605 out->area[out->data++] = 'x';
4606 out->area[out->data++] = hextab[(c >> 4) & 0xF];
4607 out->area[out->data++] = hextab[c & 0xF];
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004608 }
4609 if (buf[ptr++] == '\n') {
4610 /* we had a line break, let's return now */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004611 out->area[out->data++] = '\n';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004612 *line = ptr;
4613 return ptr;
4614 }
4615 }
4616 /* we have an incomplete line, we return it as-is */
Willy Tarreau843b7cb2018-07-13 10:54:26 +02004617 out->area[out->data++] = '\n';
Willy Tarreau97c2ae12016-11-22 18:00:20 +01004618 return ptr;
4619}
4620
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004621/* displays a <len> long memory block at <buf>, assuming first byte of <buf>
Willy Tarreaued936c52017-04-27 18:03:20 +02004622 * has address <baseaddr>. String <pfx> may be placed as a prefix in front of
4623 * each line. It may be NULL if unused. The output is emitted to file <out>.
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004624 */
Willy Tarreaued936c52017-04-27 18:03:20 +02004625void debug_hexdump(FILE *out, const char *pfx, const char *buf,
4626 unsigned int baseaddr, int len)
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004627{
Willy Tarreau73459792017-04-11 07:58:08 +02004628 unsigned int i;
4629 int b, j;
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004630
4631 for (i = 0; i < (len + (baseaddr & 15)); i += 16) {
4632 b = i - (baseaddr & 15);
Willy Tarreaued936c52017-04-27 18:03:20 +02004633 fprintf(out, "%s%08x: ", pfx ? pfx : "", i + (baseaddr & ~15));
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004634 for (j = 0; j < 8; j++) {
4635 if (b + j >= 0 && b + j < len)
4636 fprintf(out, "%02x ", (unsigned char)buf[b + j]);
4637 else
4638 fprintf(out, " ");
4639 }
4640
4641 if (b + j >= 0 && b + j < len)
4642 fputc('-', out);
4643 else
4644 fputc(' ', out);
4645
4646 for (j = 8; j < 16; j++) {
4647 if (b + j >= 0 && b + j < len)
4648 fprintf(out, " %02x", (unsigned char)buf[b + j]);
4649 else
4650 fprintf(out, " ");
4651 }
4652
4653 fprintf(out, " ");
4654 for (j = 0; j < 16; j++) {
4655 if (b + j >= 0 && b + j < len) {
4656 if (isprint((unsigned char)buf[b + j]))
4657 fputc((unsigned char)buf[b + j], out);
4658 else
4659 fputc('.', out);
4660 }
4661 else
4662 fputc(' ', out);
4663 }
4664 fputc('\n', out);
4665 }
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004666}
4667
Willy Tarreaubb869862020-04-16 10:52:41 +02004668/* Tries to report the executable path name on platforms supporting this. If
4669 * not found or not possible, returns NULL.
4670 */
4671const char *get_exec_path()
4672{
4673 const char *ret = NULL;
4674
4675#if (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 16))
4676 long execfn = getauxval(AT_EXECFN);
4677
4678 if (execfn && execfn != ENOENT)
4679 ret = (const char *)execfn;
4680#endif
4681 return ret;
4682}
4683
Baruch Siache1651b22020-07-24 07:52:20 +03004684#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreau9133e482020-03-04 10:19:36 +01004685/* calls dladdr() or dladdr1() on <addr> and <dli>. If dladdr1 is available,
4686 * also returns the symbol size in <size>, otherwise returns 0 there.
4687 */
4688static int dladdr_and_size(const void *addr, Dl_info *dli, size_t *size)
4689{
4690 int ret;
Willy Tarreau62af9c82020-03-10 07:51:48 +01004691#if (__GLIBC__ > 2 || (__GLIBC__ == 2 && __GLIBC_MINOR__ >= 3)) // most detailed one
Willy Tarreau9133e482020-03-04 10:19:36 +01004692 const ElfW(Sym) *sym;
4693
4694 ret = dladdr1(addr, dli, (void **)&sym, RTLD_DL_SYMENT);
4695 if (ret)
4696 *size = sym ? sym->st_size : 0;
4697#else
4698 ret = dladdr(addr, dli);
4699 *size = 0;
4700#endif
4701 return ret;
4702}
4703#endif
4704
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004705/* Tries to append to buffer <buf> some indications about the symbol at address
4706 * <addr> using the following form:
4707 * lib:+0xoffset (unresolvable address from lib's base)
4708 * main+0xoffset (unresolvable address from main (+/-))
4709 * lib:main+0xoffset (unresolvable lib address from main (+/-))
4710 * name (resolved exact exec address)
4711 * lib:name (resolved exact lib address)
4712 * name+0xoffset/0xsize (resolved address within exec symbol)
4713 * lib:name+0xoffset/0xsize (resolved address within lib symbol)
4714 *
4715 * The file name (lib or executable) is limited to what lies between the last
4716 * '/' and the first following '.'. An optional prefix <pfx> is prepended before
4717 * the output if not null. The file is not dumped when it's the same as the one
Baruch Siache1651b22020-07-24 07:52:20 +03004718 * that contains the "main" symbol, or when __ELF__ && USE_DL are not set.
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004719 *
4720 * The symbol's base address is returned, or NULL when unresolved, in order to
4721 * allow the caller to match it against known ones.
4722 */
Willy Tarreau45fd1032021-01-20 14:37:59 +01004723const void *resolve_sym_name(struct buffer *buf, const char *pfx, const void *addr)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004724{
4725 const struct {
4726 const void *func;
4727 const char *name;
4728 } fcts[] = {
4729 { .func = process_stream, .name = "process_stream" },
4730 { .func = task_run_applet, .name = "task_run_applet" },
4731 { .func = si_cs_io_cb, .name = "si_cs_io_cb" },
Willy Tarreau586f71b2020-12-11 15:54:36 +01004732 { .func = sock_conn_iocb, .name = "sock_conn_iocb" },
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004733 { .func = dgram_fd_handler, .name = "dgram_fd_handler" },
4734 { .func = listener_accept, .name = "listener_accept" },
Willy Tarreaud597ec22021-01-29 14:29:06 +01004735 { .func = manage_global_listener_queue, .name = "manage_global_listener_queue" },
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004736 { .func = poller_pipe_io_handler, .name = "poller_pipe_io_handler" },
4737 { .func = mworker_accept_wrapper, .name = "mworker_accept_wrapper" },
Willy Tarreau02922e12021-01-29 12:27:57 +01004738 { .func = session_expire_embryonic, .name = "session_expire_embryonic" },
Willy Tarreaufb5401f2021-01-29 12:25:23 +01004739#ifdef USE_THREAD
4740 { .func = accept_queue_process, .name = "accept_queue_process" },
4741#endif
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004742#ifdef USE_LUA
4743 { .func = hlua_process_task, .name = "hlua_process_task" },
4744#endif
Ilya Shipitsinbdec3ba2020-11-14 01:56:34 +05004745#ifdef SSL_MODE_ASYNC
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004746 { .func = ssl_async_fd_free, .name = "ssl_async_fd_free" },
4747 { .func = ssl_async_fd_handler, .name = "ssl_async_fd_handler" },
4748#endif
4749 };
4750
Baruch Siache1651b22020-07-24 07:52:20 +03004751#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004752 Dl_info dli, dli_main;
Willy Tarreau9133e482020-03-04 10:19:36 +01004753 size_t size;
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004754 const char *fname, *p;
4755#endif
4756 int i;
4757
4758 if (pfx)
4759 chunk_appendf(buf, "%s", pfx);
4760
4761 for (i = 0; i < sizeof(fcts) / sizeof(fcts[0]); i++) {
4762 if (addr == fcts[i].func) {
4763 chunk_appendf(buf, "%s", fcts[i].name);
4764 return addr;
4765 }
4766 }
4767
Baruch Siache1651b22020-07-24 07:52:20 +03004768#if (defined(__ELF__) && !defined(__linux__)) || defined(USE_DL)
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004769 /* Now let's try to be smarter */
Willy Tarreau9133e482020-03-04 10:19:36 +01004770 if (!dladdr_and_size(addr, &dli, &size))
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004771 goto unknown;
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004772
4773 /* 1. prefix the library name if it's not the same object as the one
4774 * that contains the main function. The name is picked between last '/'
4775 * and first following '.'.
4776 */
4777 if (!dladdr(main, &dli_main))
4778 dli_main.dli_fbase = NULL;
4779
4780 if (dli_main.dli_fbase != dli.dli_fbase) {
4781 fname = dli.dli_fname;
4782 p = strrchr(fname, '/');
4783 if (p++)
4784 fname = p;
4785 p = strchr(fname, '.');
4786 if (!p)
4787 p = fname + strlen(fname);
4788
4789 chunk_appendf(buf, "%.*s:", (int)(long)(p - fname), fname);
4790 }
4791
4792 /* 2. symbol name */
4793 if (dli.dli_sname) {
4794 /* known, dump it and return symbol's address (exact or relative) */
4795 chunk_appendf(buf, "%s", dli.dli_sname);
4796 if (addr != dli.dli_saddr) {
4797 chunk_appendf(buf, "+%#lx", (long)(addr - dli.dli_saddr));
Willy Tarreau9133e482020-03-04 10:19:36 +01004798 if (size)
4799 chunk_appendf(buf, "/%#lx", (long)size);
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004800 }
4801 return dli.dli_saddr;
4802 }
4803 else if (dli_main.dli_fbase != dli.dli_fbase) {
4804 /* unresolved symbol from a known library, report relative offset */
4805 chunk_appendf(buf, "+%#lx", (long)(addr - dli.dli_fbase));
4806 return NULL;
4807 }
Baruch Siache1651b22020-07-24 07:52:20 +03004808#endif /* __ELF__ && !__linux__ || USE_DL */
Willy Tarreaueb8b1ca2020-03-03 17:09:08 +01004809 unknown:
4810 /* unresolved symbol from the main file, report relative offset to main */
4811 if ((void*)addr < (void*)main)
4812 chunk_appendf(buf, "main-%#lx", (long)((void*)main - addr));
4813 else
4814 chunk_appendf(buf, "main+%#lx", (long)(addr - (void*)main));
4815 return NULL;
Willy Tarreau0ebb5112016-12-05 00:10:57 +01004816}
4817
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004818/*
4819 * Allocate an array of unsigned int with <nums> as address from <str> string
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05004820 * made of integer separated by dot characters.
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004821 *
4822 * First, initializes the value with <sz> as address to 0 and initializes the
4823 * array with <nums> as address to NULL. Then allocates the array with <nums> as
4824 * address updating <sz> pointed value to the size of this array.
4825 *
4826 * Returns 1 if succeeded, 0 if not.
4827 */
4828int parse_dotted_uints(const char *str, unsigned int **nums, size_t *sz)
4829{
4830 unsigned int *n;
4831 const char *s, *end;
4832
4833 s = str;
4834 *sz = 0;
4835 end = str + strlen(str);
4836 *nums = n = NULL;
4837
4838 while (1) {
4839 unsigned int r;
4840
4841 if (s >= end)
4842 break;
4843
4844 r = read_uint(&s, end);
4845 /* Expected characters after having read an uint: '\0' or '.',
4846 * if '.', must not be terminal.
4847 */
Christopher Faulet4b524122021-02-11 10:42:41 +01004848 if (*s != '\0'&& (*s++ != '.' || s == end)) {
4849 free(n);
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004850 return 0;
Christopher Faulet4b524122021-02-11 10:42:41 +01004851 }
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004852
Frédéric Lécaille12a71842019-02-26 18:19:48 +01004853 n = my_realloc2(n, (*sz + 1) * sizeof *n);
Frédéric Lécaille3b717162019-02-25 15:04:22 +01004854 if (!n)
4855 return 0;
4856
4857 n[(*sz)++] = r;
4858 }
4859 *nums = n;
4860
4861 return 1;
4862}
4863
Willy Tarreau4d589e72019-08-23 19:02:26 +02004864
4865/* returns the number of bytes needed to encode <v> as a varint. An inline
4866 * version exists for use with constants (__varint_bytes()).
4867 */
4868int varint_bytes(uint64_t v)
4869{
4870 int len = 1;
4871
4872 if (v >= 240) {
4873 v = (v - 240) >> 4;
4874 while (1) {
4875 len++;
4876 if (v < 128)
4877 break;
4878 v = (v - 128) >> 7;
4879 }
4880 }
4881 return len;
4882}
4883
Willy Tarreau52bf8392020-03-08 00:42:37 +01004884
4885/* Random number generator state, see below */
Willy Tarreau1544c142020-03-12 00:31:18 +01004886static uint64_t ha_random_state[2] ALIGNED(2*sizeof(uint64_t));
Willy Tarreau52bf8392020-03-08 00:42:37 +01004887
4888/* This is a thread-safe implementation of xoroshiro128** described below:
4889 * http://prng.di.unimi.it/
4890 * It features a 2^128 long sequence, returns 64 high-quality bits on each call,
4891 * supports fast jumps and passes all common quality tests. It is thread-safe,
4892 * uses a double-cas on 64-bit architectures supporting it, and falls back to a
4893 * local lock on other ones.
4894 */
4895uint64_t ha_random64()
4896{
4897 uint64_t result;
Willy Tarreau1544c142020-03-12 00:31:18 +01004898 uint64_t old[2] ALIGNED(2*sizeof(uint64_t));
4899 uint64_t new[2] ALIGNED(2*sizeof(uint64_t));
Willy Tarreau52bf8392020-03-08 00:42:37 +01004900
4901#if defined(USE_THREAD) && (!defined(HA_CAS_IS_8B) || !defined(HA_HAVE_CAS_DW))
4902 static HA_SPINLOCK_T rand_lock;
4903
4904 HA_SPIN_LOCK(OTHER_LOCK, &rand_lock);
4905#endif
4906
4907 old[0] = ha_random_state[0];
4908 old[1] = ha_random_state[1];
4909
4910#if defined(USE_THREAD) && defined(HA_CAS_IS_8B) && defined(HA_HAVE_CAS_DW)
4911 do {
4912#endif
4913 result = rotl64(old[0] * 5, 7) * 9;
4914 new[1] = old[0] ^ old[1];
4915 new[0] = rotl64(old[0], 24) ^ new[1] ^ (new[1] << 16); // a, b
4916 new[1] = rotl64(new[1], 37); // c
4917
4918#if defined(USE_THREAD) && defined(HA_CAS_IS_8B) && defined(HA_HAVE_CAS_DW)
4919 } while (unlikely(!_HA_ATOMIC_DWCAS(ha_random_state, old, new)));
4920#else
4921 ha_random_state[0] = new[0];
4922 ha_random_state[1] = new[1];
4923#if defined(USE_THREAD)
4924 HA_SPIN_UNLOCK(OTHER_LOCK, &rand_lock);
4925#endif
4926#endif
4927 return result;
4928}
4929
4930/* seeds the random state using up to <len> bytes from <seed>, starting with
4931 * the first non-zero byte.
4932 */
4933void ha_random_seed(const unsigned char *seed, size_t len)
4934{
4935 size_t pos;
4936
4937 /* the seed must not be all zeroes, so we pre-fill it with alternating
4938 * bits and overwrite part of them with the block starting at the first
4939 * non-zero byte from the seed.
4940 */
4941 memset(ha_random_state, 0x55, sizeof(ha_random_state));
4942
4943 for (pos = 0; pos < len; pos++)
4944 if (seed[pos] != 0)
4945 break;
4946
4947 if (pos == len)
4948 return;
4949
4950 seed += pos;
4951 len -= pos;
4952
4953 if (len > sizeof(ha_random_state))
4954 len = sizeof(ha_random_state);
4955
4956 memcpy(ha_random_state, seed, len);
4957}
4958
4959/* This causes a jump to (dist * 2^96) places in the pseudo-random sequence,
4960 * and is equivalent to calling ha_random64() as many times. It is used to
4961 * provide non-overlapping sequences of 2^96 numbers (~7*10^28) to up to 2^32
4962 * different generators (i.e. different processes after a fork). The <dist>
4963 * argument is the distance to jump to and is used in a loop so it rather not
4964 * be too large if the processing time is a concern.
4965 *
4966 * BEWARE: this function is NOT thread-safe and must not be called during
4967 * concurrent accesses to ha_random64().
4968 */
4969void ha_random_jump96(uint32_t dist)
4970{
4971 while (dist--) {
4972 uint64_t s0 = 0;
4973 uint64_t s1 = 0;
4974 int b;
4975
4976 for (b = 0; b < 64; b++) {
4977 if ((0xd2a98b26625eee7bULL >> b) & 1) {
4978 s0 ^= ha_random_state[0];
4979 s1 ^= ha_random_state[1];
4980 }
4981 ha_random64();
4982 }
4983
4984 for (b = 0; b < 64; b++) {
4985 if ((0xdddf9b1090aa7ac1ULL >> b) & 1) {
4986 s0 ^= ha_random_state[0];
4987 s1 ^= ha_random_state[1];
4988 }
4989 ha_random64();
4990 }
4991 ha_random_state[0] = s0;
4992 ha_random_state[1] = s1;
4993 }
4994}
4995
Willy Tarreauee3bcdd2020-03-08 17:48:17 +01004996/* Generates an RFC4122 UUID into chunk <output> which must be at least 37
4997 * bytes large.
4998 */
4999void ha_generate_uuid(struct buffer *output)
5000{
5001 uint32_t rnd[4];
5002 uint64_t last;
5003
5004 last = ha_random64();
5005 rnd[0] = last;
5006 rnd[1] = last >> 32;
5007
5008 last = ha_random64();
5009 rnd[2] = last;
5010 rnd[3] = last >> 32;
5011
5012 chunk_printf(output, "%8.8x-%4.4x-%4.4x-%4.4x-%12.12llx",
5013 rnd[0],
5014 rnd[1] & 0xFFFF,
5015 ((rnd[1] >> 16u) & 0xFFF) | 0x4000, // highest 4 bits indicate the uuid version
5016 (rnd[2] & 0x3FFF) | 0x8000, // the highest 2 bits indicate the UUID variant (10),
5017 (long long)((rnd[2] >> 14u) | ((uint64_t) rnd[3] << 18u)) & 0xFFFFFFFFFFFFull);
5018}
5019
5020
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005021/* only used by parse_line() below. It supports writing in place provided that
5022 * <in> is updated to the next location before calling it. In that case, the
5023 * char at <in> may be overwritten.
5024 */
5025#define EMIT_CHAR(x) \
5026 do { \
5027 char __c = (char)(x); \
5028 if ((opts & PARSE_OPT_INPLACE) && out+outpos > in) \
5029 err |= PARSE_ERR_OVERLAP; \
5030 if (outpos >= outmax) \
5031 err |= PARSE_ERR_TOOLARGE; \
5032 if (!err) \
5033 out[outpos] = __c; \
5034 outpos++; \
5035 } while (0)
5036
Ilya Shipitsin46a030c2020-07-05 16:36:08 +05005037/* Parse <in>, copy it into <out> split into isolated words whose pointers
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005038 * are put in <args>. If more than <outlen> bytes have to be emitted, the
5039 * extraneous ones are not emitted but <outlen> is updated so that the caller
5040 * knows how much to realloc. Similarly, <args> are not updated beyond <nbargs>
5041 * but the returned <nbargs> indicates how many were found. All trailing args
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005042 * up to <nbargs> point to the trailing zero, and as long as <nbargs> is > 0,
5043 * it is guaranteed that at least one arg will point to the zero. It is safe
5044 * to call it with a NULL <args> if <nbargs> is 0.
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005045 *
5046 * <out> may overlap with <in> provided that it never goes further, in which
5047 * case the parser will accept to perform in-place parsing and unquoting/
5048 * unescaping but only if environment variables do not lead to expansion that
5049 * causes overlapping, otherwise the input string being destroyed, the error
5050 * will not be recoverable. Note that even during out-of-place <in> will
5051 * experience temporary modifications in-place for variable resolution and must
5052 * be writable, and will also receive zeroes to delimit words when using
5053 * in-place copy. Parsing options <opts> taken from PARSE_OPT_*. Return value
5054 * is zero on success otherwise a bitwise-or of PARSE_ERR_*. Upon error, the
5055 * starting point of the first invalid character sequence or unmatched
5056 * quote/brace is reported in <errptr> if not NULL. When using in-place parsing
5057 * error reporting might be difficult since zeroes will have been inserted into
5058 * the string. One solution for the caller may consist in replacing all args
5059 * delimiters with spaces in this case.
5060 */
5061uint32_t parse_line(char *in, char *out, size_t *outlen, char **args, int *nbargs, uint32_t opts, char **errptr)
5062{
5063 char *quote = NULL;
5064 char *brace = NULL;
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005065 char *word_expand = NULL;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005066 unsigned char hex1, hex2;
5067 size_t outmax = *outlen;
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005068 int argsmax = *nbargs - 1;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005069 size_t outpos = 0;
5070 int squote = 0;
5071 int dquote = 0;
5072 int arg = 0;
5073 uint32_t err = 0;
5074
5075 *nbargs = 0;
5076 *outlen = 0;
5077
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005078 /* argsmax may be -1 here, protecting args[] from any write */
5079 if (arg < argsmax)
5080 args[arg] = out;
5081
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005082 while (1) {
5083 if (*in >= '-' && *in != '\\') {
5084 /* speedup: directly send all regular chars starting
5085 * with '-', '.', '/', alnum etc...
5086 */
5087 EMIT_CHAR(*in++);
5088 continue;
5089 }
5090 else if (*in == '\0' || *in == '\n' || *in == '\r') {
5091 /* end of line */
5092 break;
5093 }
5094 else if (*in == '#' && (opts & PARSE_OPT_SHARP) && !squote && !dquote) {
5095 /* comment */
5096 break;
5097 }
5098 else if (*in == '"' && !squote && (opts & PARSE_OPT_DQUOTE)) { /* double quote outside single quotes */
5099 if (dquote) {
5100 dquote = 0;
5101 quote = NULL;
5102 }
5103 else {
5104 dquote = 1;
5105 quote = in;
5106 }
5107 in++;
5108 continue;
5109 }
5110 else if (*in == '\'' && !dquote && (opts & PARSE_OPT_SQUOTE)) { /* single quote outside double quotes */
5111 if (squote) {
5112 squote = 0;
5113 quote = NULL;
5114 }
5115 else {
5116 squote = 1;
5117 quote = in;
5118 }
5119 in++;
5120 continue;
5121 }
5122 else if (*in == '\\' && !squote && (opts & PARSE_OPT_BKSLASH)) {
5123 /* first, we'll replace \\, \<space>, \#, \r, \n, \t, \xXX with their
5124 * C equivalent value but only when they have a special meaning and within
5125 * double quotes for some of them. Other combinations left unchanged (eg: \1).
5126 */
5127 char tosend = *in;
5128
5129 switch (in[1]) {
5130 case ' ':
5131 case '\\':
5132 tosend = in[1];
5133 in++;
5134 break;
5135
5136 case 't':
5137 tosend = '\t';
5138 in++;
5139 break;
5140
5141 case 'n':
5142 tosend = '\n';
5143 in++;
5144 break;
5145
5146 case 'r':
5147 tosend = '\r';
5148 in++;
5149 break;
5150
5151 case '#':
5152 /* escaping of "#" only if comments are supported */
5153 if (opts & PARSE_OPT_SHARP)
5154 in++;
5155 tosend = *in;
5156 break;
5157
5158 case '\'':
5159 /* escaping of "'" only outside single quotes and only if single quotes are supported */
5160 if (opts & PARSE_OPT_SQUOTE && !squote)
5161 in++;
5162 tosend = *in;
5163 break;
5164
5165 case '"':
5166 /* escaping of '"' only outside single quotes and only if double quotes are supported */
5167 if (opts & PARSE_OPT_DQUOTE && !squote)
5168 in++;
5169 tosend = *in;
5170 break;
5171
5172 case '$':
5173 /* escaping of '$' only inside double quotes and only if env supported */
5174 if (opts & PARSE_OPT_ENV && dquote)
5175 in++;
5176 tosend = *in;
5177 break;
5178
5179 case 'x':
5180 if (!ishex(in[2]) || !ishex(in[3])) {
5181 /* invalid or incomplete hex sequence */
5182 err |= PARSE_ERR_HEX;
5183 if (errptr)
5184 *errptr = in;
5185 goto leave;
5186 }
Willy Tarreauf278eec2020-07-05 21:46:32 +02005187 hex1 = toupper((unsigned char)in[2]) - '0';
5188 hex2 = toupper((unsigned char)in[3]) - '0';
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005189 if (hex1 > 9) hex1 -= 'A' - '9' - 1;
5190 if (hex2 > 9) hex2 -= 'A' - '9' - 1;
5191 tosend = (hex1 << 4) + hex2;
5192 in += 3;
5193 break;
5194
5195 default:
5196 /* other combinations are not escape sequences */
5197 break;
5198 }
5199
5200 in++;
5201 EMIT_CHAR(tosend);
5202 }
5203 else if (isspace((unsigned char)*in) && !squote && !dquote) {
5204 /* a non-escaped space is an argument separator */
5205 while (isspace((unsigned char)*in))
5206 in++;
5207 EMIT_CHAR(0);
5208 arg++;
5209 if (arg < argsmax)
5210 args[arg] = out + outpos;
5211 else
5212 err |= PARSE_ERR_TOOMANY;
5213 }
5214 else if (*in == '$' && (opts & PARSE_OPT_ENV) && (dquote || !(opts & PARSE_OPT_DQUOTE))) {
5215 /* environment variables are evaluated anywhere, or only
5216 * inside double quotes if they are supported.
5217 */
5218 char *var_name;
5219 char save_char;
5220 char *value;
5221
5222 in++;
5223
5224 if (*in == '{')
5225 brace = in++;
5226
5227 if (!isalpha((unsigned char)*in) && *in != '_') {
5228 /* unacceptable character in variable name */
5229 err |= PARSE_ERR_VARNAME;
5230 if (errptr)
5231 *errptr = in;
5232 goto leave;
5233 }
5234
5235 var_name = in;
5236 while (isalnum((unsigned char)*in) || *in == '_')
5237 in++;
5238
5239 save_char = *in;
5240 *in = '\0';
5241 value = getenv(var_name);
5242 *in = save_char;
5243
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005244 /* support for '[*]' sequence to force word expansion,
5245 * only available inside braces */
5246 if (*in == '[' && brace && (opts & PARSE_OPT_WORD_EXPAND)) {
5247 word_expand = in++;
5248
5249 if (*in++ != '*' || *in++ != ']') {
5250 err |= PARSE_ERR_WRONG_EXPAND;
5251 if (errptr)
5252 *errptr = word_expand;
5253 goto leave;
5254 }
5255 }
5256
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005257 if (brace) {
5258 if (*in != '}') {
5259 /* unmatched brace */
5260 err |= PARSE_ERR_BRACE;
5261 if (errptr)
5262 *errptr = brace;
5263 goto leave;
5264 }
5265 in++;
5266 brace = NULL;
5267 }
5268
5269 if (value) {
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005270 while (*value) {
5271 /* expand as individual parameters on a space character */
Willy Tarreaufe2cc412020-10-01 18:04:40 +02005272 if (word_expand && isspace((unsigned char)*value)) {
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005273 EMIT_CHAR(0);
5274 ++arg;
5275 if (arg < argsmax)
5276 args[arg] = out + outpos;
5277 else
5278 err |= PARSE_ERR_TOOMANY;
5279
5280 /* skip consecutive spaces */
Willy Tarreaufe2cc412020-10-01 18:04:40 +02005281 while (isspace((unsigned char)*++value))
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005282 ;
5283 } else {
5284 EMIT_CHAR(*value++);
5285 }
5286 }
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005287 }
Amaury Denoyellefa41cb62020-10-01 14:32:35 +02005288 word_expand = NULL;
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005289 }
5290 else {
5291 /* any other regular char */
5292 EMIT_CHAR(*in++);
5293 }
5294 }
5295
5296 /* end of output string */
5297 EMIT_CHAR(0);
5298 arg++;
5299
5300 if (quote) {
5301 /* unmatched quote */
5302 err |= PARSE_ERR_QUOTE;
5303 if (errptr)
5304 *errptr = quote;
5305 goto leave;
5306 }
5307 leave:
5308 *nbargs = arg;
5309 *outlen = outpos;
5310
Willy Tarreau61dd44b2020-06-25 07:35:42 +02005311 /* empty all trailing args by making them point to the trailing zero,
5312 * at least the last one in any case.
5313 */
5314 if (arg > argsmax)
5315 arg = argsmax;
5316
5317 while (arg >= 0 && arg <= argsmax)
Willy Tarreauc8d167b2020-06-16 16:27:26 +02005318 args[arg++] = out + outpos - 1;
5319
5320 return err;
5321}
5322#undef EMIT_CHAR
5323
Willy Tarreauc54e5ad2020-06-25 09:15:40 +02005324/* This is used to sanitize an input line that's about to be used for error reporting.
5325 * It will adjust <line> to print approximately <width> chars around <pos>, trying to
5326 * preserve the beginning, with leading or trailing "..." when the line is truncated.
5327 * If non-printable chars are present in the output. It returns the new offset <pos>
5328 * in the modified line. Non-printable characters are replaced with '?'. <width> must
5329 * be at least 6 to support two "..." otherwise the result is undefined. The line
5330 * itself must have at least 7 chars allocated for the same reason.
5331 */
5332size_t sanitize_for_printing(char *line, size_t pos, size_t width)
5333{
5334 size_t shift = 0;
5335 char *out = line;
5336 char *in = line;
5337 char *end = line + width;
5338
5339 if (pos >= width) {
5340 /* if we have to shift, we'll be out of context, so let's
5341 * try to put <pos> at the center of width.
5342 */
5343 shift = pos - width / 2;
5344 in += shift + 3;
5345 end = out + width - 3;
5346 out[0] = out[1] = out[2] = '.';
5347 out += 3;
5348 }
5349
5350 while (out < end && *in) {
5351 if (isspace((unsigned char)*in))
5352 *out++ = ' ';
5353 else if (isprint((unsigned char)*in))
5354 *out++ = *in;
5355 else
5356 *out++ = '?';
5357 in++;
5358 }
5359
5360 if (end < line + width) {
5361 out[0] = out[1] = out[2] = '.';
5362 out += 3;
5363 }
5364
5365 *out++ = 0;
5366 return pos - shift;
5367}
5368
Willy Tarreaubaaee002006-06-26 02:48:02 +02005369/*
5370 * Local variables:
5371 * c-indent-level: 8
5372 * c-basic-offset: 8
5373 * End:
5374 */